[Feature] add scannet instance dataset with metrics (#1230)

* add scannet instance dataset with metrics * fix import * add missing import * fix comments and docstrings * merge v1.0.0.dev0 to scannet_instance_dataset (#1) * Bump to v0.18.0 (#1148) * Update README & getting_started & changelog & version.py * resolve comments * Update the required maximum version of mmcv-full * Update changelog.md * Update getting_started.md * Add highlight in changelog.md * update imvoxelnet citation (#1153) * deprecate the support for "python setup.py test" (#1164) * add OpenMMLab website and platform links (#1176) * [Fix] Update regnet metafile (#1184) * update regnet metafile * Add Architechture in RegNet metafile * fix axis=0 rotation in master branch (#1182) * [Enhance] support semantic seg in Flip3D augmentation (#1181) * support semantic seg in flip * change box_dtype_3d to bbox3d_fields * add point cloud annotation in doc (#1185) * [Fix] Fix bug in non-distributed multi-gpu training/testing (#1197) * Fix bug in non-distributed multi-gpu training/testing * add deprecated warning to void BC-breaking * Add missing explanation of cam_intrinsic in the nuScenes dataset doc (#1193) * [Fix] Fix corner bug in different coordinates (#1212) * fix corner bug in cam coord * fix corner bugs in depth & lidar coord * Bump to v0.18.1 (#1218) * bump to v0.18.1 * fix comments * update bazel version (#1223) * [Fix] fix pointpillars bug on kitti in master branch (#1163) * fix pointpillars bug on kitti in master branch * replace with AlignedAnchorGenerator * Update the installation of MMCV (#1226) * [Fix] Fix a potential overflow bug when post_max_size is set too large in the circle_nms (#1225) * [Fix] overflow bug * [Fix] typo Co-authored-by: zeyuzeng <zeyuzeng@autox.ai> Co-authored-by: zeyu-hello <zeyuzeng@hotmail.com> * Fix PointRCNN bugs (#1224) * fix bug to allow pointnet to train in fp16 (#1207) * fix bug to allow pointnet to train in fp16 * remove unused import * fix lint * fix lint for gather_points_cuda.cu Co-authored-by: peng <maskjp@tamu.edu> * [Fix] Recheck import sorting (#1242) * fix missed mmcv * reinstall pre-commit and recheck * fix inference_demo.ipynb bug (#1236) * [Enhance] upgrade PointPillars performace on dev branch (#1166) * upgrade PointPillars performace on dev branch * update DynamicPillarFeatureNet * fix comments * change to AlignedAnchor3DRangeGenerator * change to AlignedAnchor3DRangeGenerator * fix * replace with AlignedAnchorGenerator * fix lint * update using isort * Bump to v1.0.0.rc0 (#928) * [Refactor] Main code modification for coordinate system refactor (#677) * [Enhance] Add script for data update (#774) * Fixed wrong config paths and fixed a bug in test * Fixed metafile * Coord sys refactor (main code) * Update test_waymo_dataset.py * Manually resolve conflict * Removed unused lines and fixed imports * remove coord2box and box2coord * update dir_limit_offset * Some minor improvements * Removed some \s in comments * Revert a change * Change Box3DMode to Coord3DMode where points are converted * Fix points_in_bbox function * Fix Imvoxelnet config * Revert adding a line * Fix rotation bug when batch size is 0 * Keep sign of dir_scores as before * Fix several comments * Add a comment * Fix docstring * Add data update scripts * Fix comments * fix import (#839) * [Enhance] refactor iou_neg_piecewise_sampler.py (#842) * [Refactor] Main code modification for coordinate system refactor (#677) * [Enhance] Add script for data update (#774) * Fixed wrong config paths and fixed a bug in test * Fixed metafile * Coord sys refactor (main code) * Update test_waymo_dataset.py * Manually resolve conflict * Removed unused lines and fixed imports * remove coord2box and box2coord * update dir_limit_offset * Some minor improvements * Removed some \s in comments * Revert a change * Change Box3DMode to Coord3DMode where points are converted * Fix points_in_bbox function * Fix Imvoxelnet config * Revert adding a line * Fix rotation bug when batch size is 0 * Keep sign of dir_scores as before * Fix several comments * Add a comment * Fix docstring * Add data update scripts * Fix comments * fix import * refactor iou_neg_piecewise_sampler.py * add docstring * modify docstring Co-authored-by: Yezhen Cong <52420115+THU17cyz@users.noreply.github.com> Co-authored-by: THU17cyz <congyezhen71@hotmail.com> * [Feature] Add roipooling cuda ops (#843) * [Refactor] Main code modification for coordinate system refactor (#677) * [Enhance] Add script for data update (#774) * Fixed wrong config paths and fixed a bug in test * Fixed metafile * Coord sys refactor (main code) * Update test_waymo_dataset.py * Manually resolve conflict * Removed unused lines and fixed imports * remove coord2box and box2coord * update dir_limit_offset * Some minor improvements * Removed some \s in comments * Revert a change * Change Box3DMode to Coord3DMode where points are converted * Fix points_in_bbox function * Fix Imvoxelnet config * Revert adding a line * Fix rotation bug when batch size is 0 * Keep sign of dir_scores as before * Fix several comments * Add a comment * Fix docstring * Add data update scripts * Fix comments * fix import * add roipooling cuda ops * add roi extractor * add test_roi_extractor unittest * Modify setup.py to install roipooling ops * modify docstring * remove enlarge bbox in roipoint pooling * add_roipooling_ops * modify docstring Co-authored-by: Yezhen Cong <52420115+THU17cyz@users.noreply.github.com> Co-authored-by: THU17cyz <congyezhen71@hotmail.com> * [Refactor] Refactor code structure and docstrings (#803) * refactor points_in_boxes * Merge same functions of three boxes * More docstring fixes and unify x/y/z size * Add "optional" and fix "Default" * Add "optional" and fix "Default" * Add "optional" and fix "Default" * Add "optional" and fix "Default" * Add "optional" and fix "Default" * Remove None in function param type * Fix unittest * Add comments for NMS functions * Merge methods of Points * Add unittest * Add optional and default value * Fix box conversion and add unittest * Fix comments * Add unit test * Indent * Fix CI * Remove useless \\ * Remove useless \\ * Remove useless \\ * Remove useless \\ * Remove useless \\ * Add unit test for box bev * More unit tests and refine docstrings in box_np_ops * Fix comment * Add deprecation warning * [Feature] PointXYZWHLRBBoxCoder (#856) * support PointBasedBoxCoder * fix unittest bug * support unittest in gpu * support unittest in gpu * modified docstring * add args * add args * [Enhance] Change Groupfree3D config (#855) * All mods * PointSample * PointSample * [Doc] Add tutorials/data_pipeline Chinese version (#827) * [Doc] Add tutorials/data_pipeline Chinese version * refine doc * Use the absolute link * Use the absolute link Co-authored-by: Tai-Wang <tab_wang@outlook.com> * [Doc] Add Chinese doc for `scannet_det.md` (#836) * Part * Complete * Fix comments * Fix comments * [Doc] Add Chinese doc for `waymo_det.md` (#859) * Add complete translation * Refinements * Fix comments * Fix a minor typo Co-authored-by: Tai-Wang <tab_wang@outlook.com> * Remove 2D annotations on Lyft (#867) * Add header for files (#869) * Add header for files * Add header for files * Add header for files * Add header for files * [fix] fix typos (#872) * Fix 3 unworking configs (#882) * [Fix] Fix `index.rst` for Chinese docs (#873) * Fix index.rst for zh docs * Change switch language * [Fix] Centerpoint head nested list transpose (#879) * FIX Transpose nested lists without Numpy * Removed unused Numpy import * [Enhance] Update PointFusion (#791) * update point fusion * remove LIDAR hardcode * move get_proj_mat_by_coord_type to utils * fix lint * remove todo * fix lint * [Doc] Add nuscenes_det.md Chinese version (#854) * add nus chinese doc * add nuScenes Chinese doc * fix typo * fix typo * fix typo * fix typo * fix typo * [Fix] Fix RegNet pretrained weight loading (#889) * Fix regnet pretrained weight loading * Remove unused file * Fix centerpoint tta (#892) * [Enhance] Add benchmark regression script (#808) * Initial commit * [Feature] Support DGCNN (v1.0.0.dev0) (#896) * support dgcnn * support dgcnn * support dgcnn * support dgcnn * support dgcnn * support dgcnn * support dgcnn * support dgcnn * support dgcnn * support dgcnn * fix typo * fix typo * fix typo * del gf&fa registry (wo reuse pointnet module) * fix typo * add benchmark and add copyright header (for DGCNN only) * fix typo * fix typo * fix typo * fix typo * fix typo * support dgcnn * Change cam rot_3d_in_axis (#906) * [Doc] Add coord sys tutorial pic and change links to dev branch (#912) * Modify link branch and add pic * Fix pic * Update v1.0.0rc0 changelog * Remove v1.0.0.rc0 changelog * Init v1.0.0.rc0 changelog * Fix minor typos in the zh-CN index.rst * Add master updates in changelog v1.0.0 * Update changelog_v1.0.md * Update changelog_v1.0.md * Update changelog_v1.0.md * Adjust the order of logs * Update the number of developers * Add v1.0.0.rc0 changelog * Delete changelog_v1.0.md * Add RoIPointPool3d back * Change openmmlab pre-commit hook rev * Add links for compatibility doc * Refine details * Update README.md * Update highlights * Update README_zh-CN.md * Update getting_started.md * Update getting_started.md * Update version.py * Fix the released version name * Fix version name * Update version name * Update version name * Update version name * Update version name Co-authored-by: Yezhen Cong <52420115+THU17cyz@users.noreply.github.com> Co-authored-by: Xi Liu <75658786+xiliu8006@users.noreply.github.com> Co-authored-by: THU17cyz <congyezhen71@hotmail.com> Co-authored-by: Wenhao Wu <79644370+wHao-Wu@users.noreply.github.com> Co-authored-by: dingchang <hudingchang.vendor@sensetime.com> Co-authored-by: 谢恩泽 <Johnny_ez@163.com> Co-authored-by: Robin Karlsson <34254153+robin-karlsson0@users.noreply.github.com> Co-authored-by: Danila Rukhovich <danrukh@gmail.com> Co-authored-by: ChaimZhu <zhuchenming@pjlab.org.cn> Co-authored-by: Wenhao Wu <79644370+wHao-Wu@users.noreply.github.com> Co-authored-by: ChaimZhu <zhuchenming@pjlab.org.cn> Co-authored-by: Tai-Wang <tab_wang@outlook.com> Co-authored-by: Double-Z <zengzeyu@hotmail.com> Co-authored-by: zeyuzeng <zeyuzeng@autox.ai> Co-authored-by: zeyu-hello <zeyuzeng@hotmail.com> Co-authored-by: maskjp <maskjp@163.com> Co-authored-by: peng <maskjp@tamu.edu> Co-authored-by: Yezhen Cong <52420115+THU17cyz@users.noreply.github.com> Co-authored-by: Xi Liu <75658786+xiliu8006@users.noreply.github.com> Co-authored-by: THU17cyz <congyezhen71@hotmail.com> Co-authored-by: dingchang <hudingchang.vendor@sensetime.com> Co-authored-by: 谢恩泽 <Johnny_ez@163.com> Co-authored-by: Robin Karlsson <34254153+robin-karlsson0@users.noreply.github.com> * fix lint, add docstring * V1.0.0.dev0 (#2) * Bump to v0.18.0 (#1148) * Update README & getting_started & changelog & version.py * resolve comments * Update the required maximum version of mmcv-full * Update changelog.md * Update getting_started.md * Add highlight in changelog.md * update imvoxelnet citation (#1153) * deprecate the support for "python setup.py test" (#1164) * add OpenMMLab website and platform links (#1176) * [Fix] Update regnet metafile (#1184) * update regnet metafile * Add Architechture in RegNet metafile * fix axis=0 rotation in master branch (#1182) * [Enhance] support semantic seg in Flip3D augmentation (#1181) * support semantic seg in flip * change box_dtype_3d to bbox3d_fields * add point cloud annotation in doc (#1185) * [Fix] Fix bug in non-distributed multi-gpu training/testing (#1197) * Fix bug in non-distributed multi-gpu training/testing * add deprecated warning to void BC-breaking * Add missing explanation of cam_intrinsic in the nuScenes dataset doc (#1193) * [Fix] Fix corner bug in different coordinates (#1212) * fix corner bug in cam coord * fix corner bugs in depth & lidar coord * Bump to v0.18.1 (#1218) * bump to v0.18.1 * fix comments * update bazel version (#1223) * [Fix] fix pointpillars bug on kitti in master branch (#1163) * fix pointpillars bug on kitti in master branch * replace with AlignedAnchorGenerator * Update the installation of MMCV (#1226) * [Fix] Fix a potential overflow bug when post_max_size is set too large in the circle_nms (#1225) * [Fix] overflow bug * [Fix] typo Co-authored-by: zeyuzeng <zeyuzeng@autox.ai> Co-authored-by: zeyu-hello <zeyuzeng@hotmail.com> * Fix PointRCNN bugs (#1224) * fix bug to allow pointnet to train in fp16 (#1207) * fix bug to allow pointnet to train in fp16 * remove unused import * fix lint * fix lint for gather_points_cuda.cu Co-authored-by: peng <maskjp@tamu.edu> * [Fix] Recheck import sorting (#1242) * fix missed mmcv * reinstall pre-commit and recheck * fix inference_demo.ipynb bug (#1236) * [Enhance] upgrade PointPillars performace on dev branch (#1166) * upgrade PointPillars performace on dev branch * update DynamicPillarFeatureNet * fix comments * change to AlignedAnchor3DRangeGenerator * change to AlignedAnchor3DRangeGenerator * fix * replace with AlignedAnchorGenerator * fix lint * update using isort * Bump to v1.0.0.rc0 (#928) * [Refactor] Main code modification for coordinate system refactor (#677) * [Enhance] Add script for data update (#774) * Fixed wrong config paths and fixed a bug in test * Fixed metafile * Coord sys refactor (main code) * Update test_waymo_dataset.py * Manually resolve conflict * Removed unused lines and fixed imports * remove coord2box and box2coord * update dir_limit_offset * Some minor improvements * Removed some \s in comments * Revert a change * Change Box3DMode to Coord3DMode where points are converted * Fix points_in_bbox function * Fix Imvoxelnet config * Revert adding a line * Fix rotation bug when batch size is 0 * Keep sign of dir_scores as before * Fix several comments * Add a comment * Fix docstring * Add data update scripts * Fix comments * fix import (#839) * [Enhance] refactor iou_neg_piecewise_sampler.py (#842) * [Refactor] Main code modification for coordinate system refactor (#677) * [Enhance] Add script for data update (#774) * Fixed wrong config paths and fixed a bug in test * Fixed metafile * Coord sys refactor (main code) * Update test_waymo_dataset.py * Manually resolve conflict * Removed unused lines and fixed imports * remove coord2box and box2coord * update dir_limit_offset * Some minor improvements * Removed some \s in comments * Revert a change * Change Box3DMode to Coord3DMode where points are converted * Fix points_in_bbox function * Fix Imvoxelnet config * Revert adding a line * Fix rotation bug when batch size is 0 * Keep sign of dir_scores as before * Fix several comments * Add a comment * Fix docstring * Add data update scripts * Fix comments * fix import * refactor iou_neg_piecewise_sampler.py * add docstring * modify docstring Co-authored-by: Yezhen Cong <52420115+THU17cyz@users.noreply.github.com> Co-authored-by: THU17cyz <congyezhen71@hotmail.com> * [Feature] Add roipooling cuda ops (#843) * [Refactor] Main code modification for coordinate system refactor (#677) * [Enhance] Add script for data update (#774) * Fixed wrong config paths and fixed a bug in test * Fixed metafile * Coord sys refactor (main code) * Update test_waymo_dataset.py * Manually resolve conflict * Removed unused lines and fixed imports * remove coord2box and box2coord * update dir_limit_offset * Some minor improvements * Removed some \s in comments * Revert a change * Change Box3DMode to Coord3DMode where points are converted * Fix points_in_bbox function * Fix Imvoxelnet config * Revert adding a line * Fix rotation bug when batch size is 0 * Keep sign of dir_scores as before * Fix several comments * Add a comment * Fix docstring * Add data update scripts * Fix comments * fix import * add roipooling cuda ops * add roi extractor * add test_roi_extractor unittest * Modify setup.py to install roipooling ops * modify docstring * remove enlarge bbox in roipoint pooling * add_roipooling_ops * modify docstring Co-authored-by: Yezhen Cong <52420115+THU17cyz@users.noreply.github.com> Co-authored-by: THU17cyz <congyezhen71@hotmail.com> * [Refactor] Refactor code structure and docstrings (#803) * refactor points_in_boxes * Merge same functions of three boxes * More docstring fixes and unify x/y/z size * Add "optional" and fix "Default" * Add "optional" and fix "Default" * Add "optional" and fix "Default" * Add "optional" and fix "Default" * Add "optional" and fix "Default" * Remove None in function param type * Fix unittest * Add comments for NMS functions * Merge methods of Points * Add unittest * Add optional and default value * Fix box conversion and add unittest * Fix comments * Add unit test * Indent * Fix CI * Remove useless \\ * Remove useless \\ * Remove useless \\ * Remove useless \\ * Remove useless \\ * Add unit test for box bev * More unit tests and refine docstrings in box_np_ops * Fix comment * Add deprecation warning * [Feature] PointXYZWHLRBBoxCoder (#856) * support PointBasedBoxCoder * fix unittest bug * support unittest in gpu * support unittest in gpu * modified docstring * add args * add args * [Enhance] Change Groupfree3D config (#855) * All mods * PointSample * PointSample * [Doc] Add tutorials/data_pipeline Chinese version (#827) * [Doc] Add tutorials/data_pipeline Chinese version * refine doc * Use the absolute link * Use the absolute link Co-authored-by: Tai-Wang <tab_wang@outlook.com> * [Doc] Add Chinese doc for `scannet_det.md` (#836) * Part * Complete * Fix comments * Fix comments * [Doc] Add Chinese doc for `waymo_det.md` (#859) * Add complete translation * Refinements * Fix comments * Fix a minor typo Co-authored-by: Tai-Wang <tab_wang@outlook.com> * Remove 2D annotations on Lyft (#867) * Add header for files (#869) * Add header for files * Add header for files * Add header for files * Add header for files * [fix] fix typos (#872) * Fix 3 unworking configs (#882) * [Fix] Fix `index.rst` for Chinese docs (#873) * Fix index.rst for zh docs * Change switch language * [Fix] Centerpoint head nested list transpose (#879) * FIX Transpose nested lists without Numpy * Removed unused Numpy import * [Enhance] Update PointFusion (#791) * update point fusion * remove LIDAR hardcode * move get_proj_mat_by_coord_type to utils * fix lint * remove todo * fix lint * [Doc] Add nuscenes_det.md Chinese version (#854) * add nus chinese doc * add nuScenes Chinese doc * fix typo * fix typo * fix typo * fix typo * fix typo * [Fix] Fix RegNet pretrained weight loading (#889) * Fix regnet pretrained weight loading * Remove unused file * Fix centerpoint tta (#892) * [Enhance] Add benchmark regression script (#808) * Initial commit * [Feature] Support DGCNN (v1.0.0.dev0) (#896) * support dgcnn * support dgcnn * support dgcnn * support dgcnn * support dgcnn * support dgcnn * support dgcnn * support dgcnn * support dgcnn * support dgcnn * fix typo * fix typo * fix typo * del gf&fa registry (wo reuse pointnet module) * fix typo * add benchmark and add copyright header (for DGCNN only) * fix typo * fix typo * fix typo * fix typo * fix typo * support dgcnn * Change cam rot_3d_in_axis (#906) * [Doc] Add coord sys tutorial pic and change links to dev branch (#912) * Modify link branch and add pic * Fix pic * Update v1.0.0rc0 changelog * Remove v1.0.0.rc0 changelog * Init v1.0.0.rc0 changelog * Fix minor typos in the zh-CN index.rst * Add master updates in changelog v1.0.0 * Update changelog_v1.0.md * Update changelog_v1.0.md * Update changelog_v1.0.md * Adjust the order of logs * Update the number of developers * Add v1.0.0.rc0 changelog * Delete changelog_v1.0.md * Add RoIPointPool3d back * Change openmmlab pre-commit hook rev * Add links for compatibility doc * Refine details * Update README.md * Update highlights * Update README_zh-CN.md * Update getting_started.md * Update getting_started.md * Update version.py * Fix the released version name * Fix version name * Update version name * Update version name * Update version name * Update version name Co-authored-by: Yezhen Cong <52420115+THU17cyz@users.noreply.github.com> Co-authored-by: Xi Liu <75658786+xiliu8006@users.noreply.github.com> Co-authored-by: THU17cyz <congyezhen71@hotmail.com> Co-authored-by: Wenhao Wu <79644370+wHao-Wu@users.noreply.github.com> Co-authored-by: dingchang <hudingchang.vendor@sensetime.com> Co-authored-by: 谢恩泽 <Johnny_ez@163.com> Co-authored-by: Robin Karlsson <34254153+robin-karlsson0@users.noreply.github.com> Co-authored-by: Danila Rukhovich <danrukh@gmail.com> Co-authored-by: ChaimZhu <zhuchenming@pjlab.org.cn> * fix ci in dev branch (#1264) Co-authored-by: Wenhao Wu <79644370+wHao-Wu@users.noreply.github.com> Co-authored-by: ChaimZhu <zhuchenming@pjlab.org.cn> Co-authored-by: Tai-Wang <tab_wang@outlook.com> Co-authored-by: Double-Z <zengzeyu@hotmail.com> Co-authored-by: zeyuzeng <zeyuzeng@autox.ai> Co-authored-by: zeyu-hello <zeyuzeng@hotmail.com> Co-authored-by: maskjp <maskjp@163.com> Co-authored-by: peng <maskjp@tamu.edu> Co-authored-by: Yezhen Cong <52420115+THU17cyz@users.noreply.github.com> Co-authored-by: Xi Liu <75658786+xiliu8006@users.noreply.github.com> Co-authored-by: THU17cyz <congyezhen71@hotmail.com> Co-authored-by: dingchang <hudingchang.vendor@sensetime.com> Co-authored-by: 谢恩泽 <Johnny_ez@163.com> Co-authored-by: Robin Karlsson <34254153+robin-karlsson0@users.noreply.github.com> * try revert imports Co-authored-by: Wenhao Wu <79644370+wHao-Wu@users.noreply.github.com> Co-authored-by: ChaimZhu <zhuchenming@pjlab.org.cn> Co-authored-by: Tai-Wang <tab_wang@outlook.com> Co-authored-by: Double-Z <zengzeyu@hotmail.com> Co-authored-by: zeyuzeng <zeyuzeng@autox.ai> Co-authored-by: zeyu-hello <zeyuzeng@hotmail.com> Co-authored-by: maskjp <maskjp@163.com> Co-authored-by: peng <maskjp@tamu.edu> Co-authored-by: Yezhen Cong <52420115+THU17cyz@users.noreply.github.com> Co-authored-by: Xi Liu <75658786+xiliu8006@users.noreply.github.com> Co-authored-by: THU17cyz <congyezhen71@hotmail.com> Co-authored-by: dingchang <hudingchang.vendor@sensetime.com> Co-authored-by: 谢恩泽 <Johnny_ez@163.com> Co-authored-by: Robin Karlsson <34254153+robin-karlsson0@users.noreply.github.com>

[Feature] add scannet instance dataset with metrics (#1230)
* add scannet instance dataset with metrics * fix import * add missing import * fix comments and docstrings * merge v1.0.0.dev0 to scannet_instance_dataset (#1) * Bump to v0.18.0 (#1148) * Update README & getting_started & changelog & version.py * resolve comments * Update the required maximum version of mmcv-full * Update changelog.md * Update getting_started.md * Add highlight in changelog.md * update imvoxelnet citation (#1153) * deprecate the support for "python setup.py test" (#1164) * add OpenMMLab website and platform links (#1176) * [Fix] Update regnet metafile (#1184) * update regnet metafile * Add Architechture in RegNet metafile * fix axis=0 rotation in master branch (#1182) * [Enhance] support semantic seg in Flip3D augmentation (#1181) * support semantic seg in flip * change box_dtype_3d to bbox3d_fields * add point cloud annotation in doc (#1185) * [Fix] Fix bug in non-distributed multi-gpu training/testing (#1197) * Fix bug in non-distributed multi-gpu training/testing * add deprecated warning to void BC-breaking * Add missing explanation of cam_intrinsic in the nuScenes dataset doc (#1193) * [Fix] Fix corner bug in different coordinates (#1212) * fix corner bug in cam coord * fix corner bugs in depth & lidar coord * Bump to v0.18.1 (#1218) * bump to v0.18.1 * fix comments * update bazel version (#1223) * [Fix] fix pointpillars bug on kitti in master branch (#1163) * fix pointpillars bug on kitti in master branch * replace with AlignedAnchorGenerator * Update the installation of MMCV (#1226) * [Fix] Fix a potential overflow bug when post_max_size is set too large in the circle_nms (#1225) * [Fix] overflow bug * [Fix] typo Co-authored-by: zeyuzeng <zeyuzeng@autox.ai> Co-authored-by: zeyu-hello <zeyuzeng@hotmail.com> * Fix PointRCNN bugs (#1224) * fix bug to allow pointnet to train in fp16 (#1207) * fix bug to allow pointnet to train in fp16 * remove unused import * fix lint * fix lint for gather_points_cuda.cu Co-authored-by: peng <maskjp@tamu.edu> * [Fix] Recheck import sorting (#1242) * fix missed mmcv * reinstall pre-commit and recheck * fix inference_demo.ipynb bug (#1236) * [Enhance] upgrade PointPillars performace on dev branch (#1166) * upgrade PointPillars performace on dev branch * update DynamicPillarFeatureNet * fix comments * change to AlignedAnchor3DRangeGenerator * change to AlignedAnchor3DRangeGenerator * fix * replace with AlignedAnchorGenerator * fix lint * update using isort * Bump to v1.0.0.rc0 (#928) * [Refactor] Main code modification for coordinate system refactor (#677) * [Enhance] Add script for data update (#774) * Fixed wrong config paths and fixed a bug in test * Fixed metafile * Coord sys refactor (main code) * Update test_waymo_dataset.py * Manually resolve conflict * Removed unused lines and fixed imports * remove coord2box and box2coord * update dir_limit_offset * Some minor improvements * Removed some \s in comments * Revert a change * Change Box3DMode to Coord3DMode where points are converted * Fix points_in_bbox function * Fix Imvoxelnet config * Revert adding a line * Fix rotation bug when batch size is 0 * Keep sign of dir_scores as before * Fix several comments * Add a comment * Fix docstring * Add data update scripts * Fix comments * fix import (#839) * [Enhance] refactor iou_neg_piecewise_sampler.py (#842) * [Refactor] Main code modification for coordinate system refactor (#677) * [Enhance] Add script for data update (#774) * Fixed wrong config paths and fixed a bug in test * Fixed metafile * Coord sys refactor (main code) * Update test_waymo_dataset.py * Manually resolve conflict * Removed unused lines and fixed imports * remove coord2box and box2coord * update dir_limit_offset * Some minor improvements * Removed some \s in comments * Revert a change * Change Box3DMode to Coord3DMode where points are converted * Fix points_in_bbox function * Fix Imvoxelnet config * Revert adding a line * Fix rotation bug when batch size is 0 * Keep sign of dir_scores as before * Fix several comments * Add a comment * Fix docstring * Add data update scripts * Fix comments * fix import * refactor iou_neg_piecewise_sampler.py * add docstring * modify docstring Co-authored-by: Yezhen Cong <52420115+THU17cyz@users.noreply.github.com> Co-authored-by: THU17cyz <congyezhen71@hotmail.com> * [Feature] Add roipooling cuda ops (#843) * [Refactor] Main code modification for coordinate system refactor (#677) * [Enhance] Add script for data update (#774) * Fixed wrong config paths and fixed a bug in test * Fixed metafile * Coord sys refactor (main code) * Update test_waymo_dataset.py * Manually resolve conflict * Removed unused lines and fixed imports * remove coord2box and box2coord * update dir_limit_offset * Some minor improvements * Removed some \s in comments * Revert a change * Change Box3DMode to Coord3DMode where points are converted * Fix points_in_bbox function * Fix Imvoxelnet config * Revert adding a line * Fix rotation bug when batch size is 0 * Keep sign of dir_scores as before * Fix several comments * Add a comment * Fix docstring * Add data update scripts * Fix comments * fix import * add roipooling cuda ops * add roi extractor * add test_roi_extractor unittest * Modify setup.py to install roipooling ops * modify docstring * remove enlarge bbox in roipoint pooling * add_roipooling_ops * modify docstring Co-authored-by: Yezhen Cong <52420115+THU17cyz@users.noreply.github.com> Co-authored-by: THU17cyz <congyezhen71@hotmail.com> * [Refactor] Refactor code structure and docstrings (#803) * refactor points_in_boxes * Merge same functions of three boxes * More docstring fixes and unify x/y/z size * Add "optional" and fix "Default" * Add "optional" and fix "Default" * Add "optional" and fix "Default" * Add "optional" and fix "Default" * Add "optional" and fix "Default" * Remove None in function param type * Fix unittest * Add comments for NMS functions * Merge methods of Points * Add unittest * Add optional and default value * Fix box conversion and add unittest * Fix comments * Add unit test * Indent * Fix CI * Remove useless \\ * Remove useless \\ * Remove useless \\ * Remove useless \\ * Remove useless \\ * Add unit test for box bev * More unit tests and refine docstrings in box_np_ops * Fix comment * Add deprecation warning * [Feature] PointXYZWHLRBBoxCoder (#856) * support PointBasedBoxCoder * fix unittest bug * support unittest in gpu * support unittest in gpu * modified docstring * add args * add args * [Enhance] Change Groupfree3D config (#855) * All mods * PointSample * PointSample * [Doc] Add tutorials/data_pipeline Chinese version (#827) * [Doc] Add tutorials/data_pipeline Chinese version * refine doc * Use the absolute link * Use the absolute link Co-authored-by: Tai-Wang <tab_wang@outlook.com> * [Doc] Add Chinese doc for `scannet_det.md` (#836) * Part * Complete * Fix comments * Fix comments * [Doc] Add Chinese doc for `waymo_det.md` (#859) * Add complete translation * Refinements * Fix comments * Fix a minor typo Co-authored-by: Tai-Wang <tab_wang@outlook.com> * Remove 2D annotations on Lyft (#867) * Add header for files (#869) * Add header for files * Add header for files * Add header for files * Add header for files * [fix] fix typos (#872) * Fix 3 unworking configs (#882) * [Fix] Fix `index.rst` for Chinese docs (#873) * Fix index.rst for zh docs * Change switch language * [Fix] Centerpoint head nested list transpose (#879) * FIX Transpose nested lists without Numpy * Removed unused Numpy import * [Enhance] Update PointFusion (#791) * update point fusion * remove LIDAR hardcode * move get_proj_mat_by_coord_type to utils * fix lint * remove todo * fix lint * [Doc] Add nuscenes_det.md Chinese version (#854) * add nus chinese doc * add nuScenes Chinese doc * fix typo * fix typo * fix typo * fix typo * fix typo * [Fix] Fix RegNet pretrained weight loading (#889) * Fix regnet pretrained weight loading * Remove unused file * Fix centerpoint tta (#892) * [Enhance] Add benchmark regression script (#808) * Initial commit * [Feature] Support DGCNN (v1.0.0.dev0) (#896) * support dgcnn * support dgcnn * support dgcnn * support dgcnn * support dgcnn * support dgcnn * support dgcnn * support dgcnn * support dgcnn * support dgcnn * fix typo * fix typo * fix typo * del gf&fa registry (wo reuse pointnet module) * fix typo * add benchmark and add copyright header (for DGCNN only) * fix typo * fix typo * fix typo * fix typo * fix typo * support dgcnn * Change cam rot_3d_in_axis (#906) * [Doc] Add coord sys tutorial pic and change links to dev branch (#912) * Modify link branch and add pic * Fix pic * Update v1.0.0rc0 changelog * Remove v1.0.0.rc0 changelog * Init v1.0.0.rc0 changelog * Fix minor typos in the zh-CN index.rst * Add master updates in changelog v1.0.0 * Update changelog_v1.0.md * Update changelog_v1.0.md * Update changelog_v1.0.md * Adjust the order of logs * Update the number of developers * Add v1.0.0.rc0 changelog * Delete changelog_v1.0.md * Add RoIPointPool3d back * Change openmmlab pre-commit hook rev * Add links for compatibility doc * Refine details * Update README.md * Update highlights * Update README_zh-CN.md * Update getting_started.md * Update getting_started.md * Update version.py * Fix the released version name * Fix version name * Update version name * Update version name * Update version name * Update version name Co-authored-by: Yezhen Cong <52420115+THU17cyz@users.noreply.github.com> Co-authored-by: Xi Liu <75658786+xiliu8006@users.noreply.github.com> Co-authored-by: THU17cyz <congyezhen71@hotmail.com> Co-authored-by: Wenhao Wu <79644370+wHao-Wu@users.noreply.github.com> Co-authored-by: dingchang <hudingchang.vendor@sensetime.com> Co-authored-by: 谢恩泽 <Johnny_ez@163.com> Co-authored-by: Robin Karlsson <34254153+robin-karlsson0@users.noreply.github.com> Co-authored-by: Danila Rukhovich <danrukh@gmail.com> Co-authored-by: ChaimZhu <zhuchenming@pjlab.org.cn> Co-authored-by: Wenhao Wu <79644370+wHao-Wu@users.noreply.github.com> Co-authored-by: ChaimZhu <zhuchenming@pjlab.org.cn> Co-authored-by: Tai-Wang <tab_wang@outlook.com> Co-authored-by: Double-Z <zengzeyu@hotmail.com> Co-authored-by: zeyuzeng <zeyuzeng@autox.ai> Co-authored-by: zeyu-hello <zeyuzeng@hotmail.com> Co-authored-by: maskjp <maskjp@163.com> Co-authored-by: peng <maskjp@tamu.edu> Co-authored-by: Yezhen Cong <52420115+THU17cyz@users.noreply.github.com> Co-authored-by: Xi Liu <75658786+xiliu8006@users.noreply.github.com> Co-authored-by: THU17cyz <congyezhen71@hotmail.com> Co-authored-by: dingchang <hudingchang.vendor@sensetime.com> Co-authored-by: 谢恩泽 <Johnny_ez@163.com> Co-authored-by: Robin Karlsson <34254153+robin-karlsson0@users.noreply.github.com> * fix lint, add docstring * V1.0.0.dev0 (#2) * Bump to v0.18.0 (#1148) * Update README & getting_started & changelog & version.py * resolve comments * Update the required maximum version of mmcv-full * Update changelog.md * Update getting_started.md * Add highlight in changelog.md * update imvoxelnet citation (#1153) * deprecate the support for "python setup.py test" (#1164) * add OpenMMLab website and platform links (#1176) * [Fix] Update regnet metafile (#1184) * update regnet metafile * Add Architechture in RegNet metafile * fix axis=0 rotation in master branch (#1182) * [Enhance] support semantic seg in Flip3D augmentation (#1181) * support semantic seg in flip * change box_dtype_3d to bbox3d_fields * add point cloud annotation in doc (#1185) * [Fix] Fix bug in non-distributed multi-gpu training/testing (#1197) * Fix bug in non-distributed multi-gpu training/testing * add deprecated warning to void BC-breaking * Add missing explanation of cam_intrinsic in the nuScenes dataset doc (#1193) * [Fix] Fix corner bug in different coordinates (#1212) * fix corner bug in cam coord * fix corner bugs in depth & lidar coord * Bump to v0.18.1 (#1218) * bump to v0.18.1 * fix comments * update bazel version (#1223) * [Fix] fix pointpillars bug on kitti in master branch (#1163) * fix pointpillars bug on kitti in master branch * replace with AlignedAnchorGenerator * Update the installation of MMCV (#1226) * [Fix] Fix a potential overflow bug when post_max_size is set too large in the circle_nms (#1225) * [Fix] overflow bug * [Fix] typo Co-authored-by: zeyuzeng <zeyuzeng@autox.ai> Co-authored-by: zeyu-hello <zeyuzeng@hotmail.com> * Fix PointRCNN bugs (#1224) * fix bug to allow pointnet to train in fp16 (#1207) * fix bug to allow pointnet to train in fp16 * remove unused import * fix lint * fix lint for gather_points_cuda.cu Co-authored-by: peng <maskjp@tamu.edu> * [Fix] Recheck import sorting (#1242) * fix missed mmcv * reinstall pre-commit and recheck * fix inference_demo.ipynb bug (#1236) * [Enhance] upgrade PointPillars performace on dev branch (#1166) * upgrade PointPillars performace on dev branch * update DynamicPillarFeatureNet * fix comments * change to AlignedAnchor3DRangeGenerator * change to AlignedAnchor3DRangeGenerator * fix * replace with AlignedAnchorGenerator * fix lint * update using isort * Bump to v1.0.0.rc0 (#928) * [Refactor] Main code modification for coordinate system refactor (#677) * [Enhance] Add script for data update (#774) * Fixed wrong config paths and fixed a bug in test * Fixed metafile * Coord sys refactor (main code) * Update test_waymo_dataset.py * Manually resolve conflict * Removed unused lines and fixed imports * remove coord2box and box2coord * update dir_limit_offset * Some minor improvements * Removed some \s in comments * Revert a change * Change Box3DMode to Coord3DMode where points are converted * Fix points_in_bbox function * Fix Imvoxelnet config * Revert adding a line * Fix rotation bug when batch size is 0 * Keep sign of dir_scores as before * Fix several comments * Add a comment * Fix docstring * Add data update scripts * Fix comments * fix import (#839) * [Enhance] refactor iou_neg_piecewise_sampler.py (#842) * [Refactor] Main code modification for coordinate system refactor (#677) * [Enhance] Add script for data update (#774) * Fixed wrong config paths and fixed a bug in test * Fixed metafile * Coord sys refactor (main code) * Update test_waymo_dataset.py * Manually resolve conflict * Removed unused lines and fixed imports * remove coord2box and box2coord * update dir_limit_offset * Some minor improvements * Removed some \s in comments * Revert a change * Change Box3DMode to Coord3DMode where points are converted * Fix points_in_bbox function * Fix Imvoxelnet config * Revert adding a line * Fix rotation bug when batch size is 0 * Keep sign of dir_scores as before * Fix several comments * Add a comment * Fix docstring * Add data update scripts * Fix comments * fix import * refactor iou_neg_piecewise_sampler.py * add docstring * modify docstring Co-authored-by: Yezhen Cong <52420115+THU17cyz@users.noreply.github.com> Co-authored-by: THU17cyz <congyezhen71@hotmail.com> * [Feature] Add roipooling cuda ops (#843) * [Refactor] Main code modification for coordinate system refactor (#677) * [Enhance] Add script for data update (#774) * Fixed wrong config paths and fixed a bug in test * Fixed metafile * Coord sys refactor (main code) * Update test_waymo_dataset.py * Manually resolve conflict * Removed unused lines and fixed imports * remove coord2box and box2coord * update dir_limit_offset * Some minor improvements * Removed some \s in comments * Revert a change * Change Box3DMode to Coord3DMode where points are converted * Fix points_in_bbox function * Fix Imvoxelnet config * Revert adding a line * Fix rotation bug when batch size is 0 * Keep sign of dir_scores as before * Fix several comments * Add a comment * Fix docstring * Add data update scripts * Fix comments * fix import * add roipooling cuda ops * add roi extractor * add test_roi_extractor unittest * Modify setup.py to install roipooling ops * modify docstring * remove enlarge bbox in roipoint pooling * add_roipooling_ops * modify docstring Co-authored-by: Yezhen Cong <52420115+THU17cyz@users.noreply.github.com> Co-authored-by: THU17cyz <congyezhen71@hotmail.com> * [Refactor] Refactor code structure and docstrings (#803) * refactor points_in_boxes * Merge same functions of three boxes * More docstring fixes and unify x/y/z size * Add "optional" and fix "Default" * Add "optional" and fix "Default" * Add "optional" and fix "Default" * Add "optional" and fix "Default" * Add "optional" and fix "Default" * Remove None in function param type * Fix unittest * Add comments for NMS functions * Merge methods of Points * Add unittest * Add optional and default value * Fix box conversion and add unittest * Fix comments * Add unit test * Indent * Fix CI * Remove useless \\ * Remove useless \\ * Remove useless \\ * Remove useless \\ * Remove useless \\ * Add unit test for box bev * More unit tests and refine docstrings in box_np_ops * Fix comment * Add deprecation warning * [Feature] PointXYZWHLRBBoxCoder (#856) * support PointBasedBoxCoder * fix unittest bug * support unittest in gpu * support unittest in gpu * modified docstring * add args * add args * [Enhance] Change Groupfree3D config (#855) * All mods * PointSample * PointSample * [Doc] Add tutorials/data_pipeline Chinese version (#827) * [Doc] Add tutorials/data_pipeline Chinese version * refine doc * Use the absolute link * Use the absolute link Co-authored-by: Tai-Wang <tab_wang@outlook.com> * [Doc] Add Chinese doc for `scannet_det.md` (#836) * Part * Complete * Fix comments * Fix comments * [Doc] Add Chinese doc for `waymo_det.md` (#859) * Add complete translation * Refinements * Fix comments * Fix a minor typo Co-authored-by: Tai-Wang <tab_wang@outlook.com> * Remove 2D annotations on Lyft (#867) * Add header for files (#869) * Add header for files * Add header for files * Add header for files * Add header for files * [fix] fix typos (#872) * Fix 3 unworking configs (#882) * [Fix] Fix `index.rst` for Chinese docs (#873) * Fix index.rst for zh docs * Change switch language * [Fix] Centerpoint head nested list transpose (#879) * FIX Transpose nested lists without Numpy * Removed unused Numpy import * [Enhance] Update PointFusion (#791) * update point fusion * remove LIDAR hardcode * move get_proj_mat_by_coord_type to utils * fix lint * remove todo * fix lint * [Doc] Add nuscenes_det.md Chinese version (#854) * add nus chinese doc * add nuScenes Chinese doc * fix typo * fix typo * fix typo * fix typo * fix typo * [Fix] Fix RegNet pretrained weight loading (#889) * Fix regnet pretrained weight loading * Remove unused file * Fix centerpoint tta (#892) * [Enhance] Add benchmark regression script (#808) * Initial commit * [Feature] Support DGCNN (v1.0.0.dev0) (#896) * support dgcnn * support dgcnn * support dgcnn * support dgcnn * support dgcnn * support dgcnn * support dgcnn * support dgcnn * support dgcnn * support dgcnn * fix typo * fix typo * fix typo * del gf&fa registry (wo reuse pointnet module) * fix typo * add benchmark and add copyright header (for DGCNN only) * fix typo * fix typo * fix typo * fix typo * fix typo * support dgcnn * Change cam rot_3d_in_axis (#906) * [Doc] Add coord sys tutorial pic and change links to dev branch (#912) * Modify link branch and add pic * Fix pic * Update v1.0.0rc0 changelog * Remove v1.0.0.rc0 changelog * Init v1.0.0.rc0 changelog * Fix minor typos in the zh-CN index.rst * Add master updates in changelog v1.0.0 * Update changelog_v1.0.md * Update changelog_v1.0.md * Update changelog_v1.0.md * Adjust the order of logs * Update the number of developers * Add v1.0.0.rc0 changelog * Delete changelog_v1.0.md * Add RoIPointPool3d back * Change openmmlab pre-commit hook rev * Add links for compatibility doc * Refine details * Update README.md * Update highlights * Update README_zh-CN.md * Update getting_started.md * Update getting_started.md * Update version.py * Fix the released version name * Fix version name * Update version name * Update version name * Update version name * Update version name Co-authored-by: Yezhen Cong <52420115+THU17cyz@users.noreply.github.com> Co-authored-by: Xi Liu <75658786+xiliu8006@users.noreply.github.com> Co-authored-by: THU17cyz <congyezhen71@hotmail.com> Co-authored-by: Wenhao Wu <79644370+wHao-Wu@users.noreply.github.com> Co-authored-by: dingchang <hudingchang.vendor@sensetime.com> Co-authored-by: 谢恩泽 <Johnny_ez@163.com> Co-authored-by: Robin Karlsson <34254153+robin-karlsson0@users.noreply.github.com> Co-authored-by: Danila Rukhovich <danrukh@gmail.com> Co-authored-by: ChaimZhu <zhuchenming@pjlab.org.cn> * fix ci in dev branch (#1264) Co-authored-by: Wenhao Wu <79644370+wHao-Wu@users.noreply.github.com> Co-authored-by: ChaimZhu <zhuchenming@pjlab.org.cn> Co-authored-by: Tai-Wang <tab_wang@outlook.com> Co-authored-by: Double-Z <zengzeyu@hotmail.com> Co-authored-by: zeyuzeng <zeyuzeng@autox.ai> Co-authored-by: zeyu-hello <zeyuzeng@hotmail.com> Co-authored-by: maskjp <maskjp@163.com> Co-authored-by: peng <maskjp@tamu.edu> Co-authored-by: Yezhen Cong <52420115+THU17cyz@users.noreply.github.com> Co-authored-by: Xi Liu <75658786+xiliu8006@users.noreply.github.com> Co-authored-by: THU17cyz <congyezhen71@hotmail.com> Co-authored-by: dingchang <hudingchang.vendor@sensetime.com> Co-authored-by: 谢恩泽 <Johnny_ez@163.com> Co-authored-by: Robin Karlsson <34254153+robin-karlsson0@users.noreply.github.com> * try revert imports Co-authored-by: Wenhao Wu <79644370+wHao-Wu@users.noreply.github.com> Co-authored-by: ChaimZhu <zhuchenming@pjlab.org.cn> Co-authored-by: Tai-Wang <tab_wang@outlook.com> Co-authored-by: Double-Z <zengzeyu@hotmail.com> Co-authored-by: zeyuzeng <zeyuzeng@autox.ai> Co-authored-by: zeyu-hello <zeyuzeng@hotmail.com> Co-authored-by: maskjp <maskjp@163.com> Co-authored-by: peng <maskjp@tamu.edu> Co-authored-by: Yezhen Cong <52420115+THU17cyz@users.noreply.github.com> Co-authored-by: Xi Liu <75658786+xiliu8006@users.noreply.github.com> Co-authored-by: THU17cyz <congyezhen71@hotmail.com> Co-authored-by: dingchang <hudingchang.vendor@sensetime.com> Co-authored-by: 谢恩泽 <Johnny_ez@163.com> Co-authored-by: Robin Karlsson <34254153+robin-karlsson0@users.noreply.github.com>
e3f0e8b0 · Danila Rukhovich · GitHub · e8fe89f5 · e3f0e8b0 · e3f0e8b0
Unverified Commit e3f0e8b0 authored Mar 09, 2022 by Danila Rukhovich Committed by GitHub Mar 09, 2022
8 changed files
--- a/mmdet3d/core/evaluation/__init__.py
+++ b/mmdet3d/core/evaluation/__init__.py
 # Copyright (c) OpenMMLab. All rights reserved.
 from .indoor_eval import indoor_eval
+from .instance_seg_eval import instance_seg_eval
 from .kitti_utils import kitti_eval, kitti_eval_coco_style
 from .lyft_eval import lyft_eval
 from .seg_eval import seg_eval

 __all__ = [
    'kitti_eval_coco_style', 'kitti_eval', 'indoor_eval', 'lyft_eval',
-    'seg_eval'
+    'seg_eval', 'instance_seg_eval'
 ]
--- a/mmdet3d/core/evaluation/instance_seg_eval.py
+++ b/mmdet3d/core/evaluation/instance_seg_eval.py
+# Copyright (c) OpenMMLab. All rights reserved.
+import numpy as np
+from mmcv.utils import print_log
+from terminaltables import AsciiTable
+
+from .scannet_utils.evaluate_semantic_instance import scannet_eval
+
+
+def aggregate_predictions(masks, labels, scores, valid_class_ids):
+    """Maps predictions to ScanNet evaluator format.
+
+    Args:
+        masks (list[torch.Tensor]): Per scene predicted instance masks.
+        labels (list[torch.Tensor]): Per scene predicted instance labels.
+        scores (list[torch.Tensor]): Per scene predicted instance scores.
+        valid_class_ids (tuple[int]): Ids of valid categories.
+
+    Returns:
+        list[dict]: Per scene aggregated predictions.
+    """
+    infos = []
+    for id, (mask, label, score) in enumerate(zip(masks, labels, scores)):
+        mask = mask.clone().numpy()
+        label = label.clone().numpy()
+        score = score.clone().numpy()
+        info = dict()
+        n_instances = mask.max() + 1
+        for i in range(n_instances):
+            # match pred_instance['filename'] from assign_instances_for_scan
+            file_name = f'{id}_{i}'
+            info[file_name] = dict()
+            info[file_name]['mask'] = (mask == i).astype(np.int)
+            info[file_name]['label_id'] = valid_class_ids[label[i]]
+            info[file_name]['conf'] = score[i]
+        infos.append(info)
+    return infos
+
+
+def rename_gt(gt_semantic_masks, gt_instance_masks, valid_class_ids):
+    """Maps gt instance and semantic masks to instance masks for ScanNet
+    evaluator.
+
+    Args:
+        gt_semantic_masks (list[torch.Tensor]): Per scene gt semantic masks.
+        gt_instance_masks (list[torch.Tensor]): Per scene gt instance masks.
+        valid_class_ids (tuple[int]): Ids of valid categories.
+
+    Returns:
+        list[np.array]: Per scene instance masks.
+    """
+    renamed_instance_masks = []
+    for semantic_mask, instance_mask in zip(gt_semantic_masks,
+                                            gt_instance_masks):
+        semantic_mask = semantic_mask.clone().numpy()
+        instance_mask = instance_mask.clone().numpy()
+        unique = np.unique(instance_mask)
+        assert len(unique) < 1000
+        for i in unique:
+            semantic_instance = semantic_mask[instance_mask == i]
+            semantic_unique = np.unique(semantic_instance)
+            assert len(semantic_unique) == 1
+            if semantic_unique[0] < len(valid_class_ids):
+                instance_mask[
+                    instance_mask ==
+                    i] = 1000 * valid_class_ids[semantic_unique[0]] + i
+        renamed_instance_masks.append(instance_mask)
+    return renamed_instance_masks
+
+
+def instance_seg_eval(gt_semantic_masks,
+                      gt_instance_masks,
+                      pred_instance_masks,
+                      pred_instance_labels,
+                      pred_instance_scores,
+                      valid_class_ids,
+                      class_labels,
+                      options=None,
+                      logger=None):
+    """Instance Segmentation Evaluation.
+
+    Evaluate the result of the instance segmentation.
+
+    Args:
+        gt_semantic_masks (list[torch.Tensor]): Ground truth semantic masks.
+        gt_instance_masks (list[torch.Tensor]): Ground truth instance masks.
+        pred_instance_masks (list[torch.Tensor]): Predicted instance masks.
+        pred_instance_labels (list[torch.Tensor]): Predicted instance labels.
+        pred_instance_scores (list[torch.Tensor]): Predicted instance labels.
+        valid_class_ids (tuple[int]): Ids of valid categories.
+        class_labels (tuple[str]): Names of valid categories.
+        options (dict, optional): Additional options. Keys may contain:
+            `overlaps`, `min_region_sizes`, `distance_threshes`,
+            `distance_confs`. Default: None.
+        logger (logging.Logger | str, optional): The way to print the mAP
+            summary. See `mmdet.utils.print_log()` for details. Default: None.
+
+    Returns:
+        dict[str, float]: Dict of results.
+    """
+    assert len(valid_class_ids) == len(class_labels)
+    id_to_label = {
+        valid_class_ids[i]: class_labels[i]
+        for i in range(len(valid_class_ids))
+    }
+    preds = aggregate_predictions(
+        masks=pred_instance_masks,
+        labels=pred_instance_labels,
+        scores=pred_instance_scores,
+        valid_class_ids=valid_class_ids)
+    gts = rename_gt(gt_semantic_masks, gt_instance_masks, valid_class_ids)
+    metrics = scannet_eval(
+        preds=preds,
+        gts=gts,
+        options=options,
+        valid_class_ids=valid_class_ids,
+        class_labels=class_labels,
+        id_to_label=id_to_label)
+    header = ['classes', 'AP_0.25', 'AP_0.50', 'AP']
+    rows = []
+    for label, data in metrics['classes'].items():
+        aps = [data['ap25%'], data['ap50%'], data['ap']]
+        rows.append([label] + [f'{ap:.4f}' for ap in aps])
+    aps = metrics['all_ap_25%'], metrics['all_ap_50%'], metrics['all_ap']
+    footer = ['Overall'] + [f'{ap:.4f}' for ap in aps]
+    table = AsciiTable([header] + rows + [footer])
+    table.inner_footing_row_border = True
+    print_log('\n' + table.table, logger=logger)
+    return metrics
--- a/mmdet3d/core/evaluation/scannet_utils/evaluate_semantic_instance.py
+++ b/mmdet3d/core/evaluation/scannet_utils/evaluate_semantic_instance.py
+# Copyright (c) OpenMMLab. All rights reserved.
+# adapted from https://github.com/ScanNet/ScanNet/blob/master/BenchmarkScripts/3d_evaluation/evaluate_semantic_instance.py # noqa
+from copy import deepcopy
+
+import numpy as np
+
+from . import util_3d
+
+
+def evaluate_matches(matches, class_labels, options):
+    """Evaluate instance segmentation from matched gt and predicted instances
+    for all scenes.
+
+    Args:
+        matches (dict): Contains gt2pred and pred2gt infos for every scene.
+        class_labels (tuple[str]): Class names.
+        options (dict): ScanNet evaluator options. See get_options.
+
+    Returns:
+        np.array: Average precision scores for all thresholds and categories.
+    """
+    overlaps = options['overlaps']
+    min_region_sizes = [options['min_region_sizes'][0]]
+    dist_threshes = [options['distance_threshes'][0]]
+    dist_confs = [options['distance_confs'][0]]
+
+    # results: class x overlap
+    ap = np.zeros((len(dist_threshes), len(class_labels), len(overlaps)),
+                  np.float)
+    for di, (min_region_size, distance_thresh, distance_conf) in enumerate(
+            zip(min_region_sizes, dist_threshes, dist_confs)):
+        for oi, overlap_th in enumerate(overlaps):
+            pred_visited = {}
+            for m in matches:
+                for label_name in class_labels:
+                    for p in matches[m]['pred'][label_name]:
+                        if 'filename' in p:
+                            pred_visited[p['filename']] = False
+            for li, label_name in enumerate(class_labels):
+                y_true = np.empty(0)
+                y_score = np.empty(0)
+                hard_false_negatives = 0
+                has_gt = False
+                has_pred = False
+                for m in matches:
+                    pred_instances = matches[m]['pred'][label_name]
+                    gt_instances = matches[m]['gt'][label_name]
+                    # filter groups in ground truth
+                    gt_instances = [
+                        gt for gt in gt_instances
+                        if gt['instance_id'] >= 1000 and gt['vert_count'] >=
+                        min_region_size and gt['med_dist'] <= distance_thresh
+                        and gt['dist_conf'] >= distance_conf
+                    ]
+                    if gt_instances:
+                        has_gt = True
+                    if pred_instances:
+                        has_pred = True
+
+                    cur_true = np.ones(len(gt_instances))
+                    cur_score = np.ones(len(gt_instances)) * (-float('inf'))
+                    cur_match = np.zeros(len(gt_instances), dtype=np.bool)
+                    # collect matches
+                    for (gti, gt) in enumerate(gt_instances):
+                        found_match = False
+                        for pred in gt['matched_pred']:
+                            # greedy assignments
+                            if pred_visited[pred['filename']]:
+                                continue
+                            overlap = float(pred['intersection']) / (
+                                gt['vert_count'] + pred['vert_count'] -
+                                pred['intersection'])
+                            if overlap > overlap_th:
+                                confidence = pred['confidence']
+                                # if already have a prediction for this gt,
+                                # the prediction with the lower score is automatically a false positive # noqa
+                                if cur_match[gti]:
+                                    max_score = max(cur_score[gti], confidence)
+                                    min_score = min(cur_score[gti], confidence)
+                                    cur_score[gti] = max_score
+                                    # append false positive
+                                    cur_true = np.append(cur_true, 0)
+                                    cur_score = np.append(cur_score, min_score)
+                                    cur_match = np.append(cur_match, True)
+                                # otherwise set score
+                                else:
+                                    found_match = True
+                                    cur_match[gti] = True
+                                    cur_score[gti] = confidence
+                                    pred_visited[pred['filename']] = True
+                        if not found_match:
+                            hard_false_negatives += 1
+                    # remove non-matched ground truth instances
+                    cur_true = cur_true[cur_match]
+                    cur_score = cur_score[cur_match]
+
+                    # collect non-matched predictions as false positive
+                    for pred in pred_instances:
+                        found_gt = False
+                        for gt in pred['matched_gt']:
+                            overlap = float(gt['intersection']) / (
+                                gt['vert_count'] + pred['vert_count'] -
+                                gt['intersection'])
+                            if overlap > overlap_th:
+                                found_gt = True
+                                break
+                        if not found_gt:
+                            num_ignore = pred['void_intersection']
+                            for gt in pred['matched_gt']:
+                                # group?
+                                if gt['instance_id'] < 1000:
+                                    num_ignore += gt['intersection']
+                                # small ground truth instances
+                                if gt['vert_count'] < min_region_size or gt[
+                                        'med_dist'] > distance_thresh or gt[
+                                            'dist_conf'] < distance_conf:
+                                    num_ignore += gt['intersection']
+                            proportion_ignore = float(
+                                num_ignore) / pred['vert_count']
+                            # if not ignored append false positive
+                            if proportion_ignore <= overlap_th:
+                                cur_true = np.append(cur_true, 0)
+                                confidence = pred['confidence']
+                                cur_score = np.append(cur_score, confidence)
+
+                    # append to overall results
+                    y_true = np.append(y_true, cur_true)
+                    y_score = np.append(y_score, cur_score)
+
+                # compute average precision
+                if has_gt and has_pred:
+                    # compute precision recall curve first
+
+                    # sorting and cumsum
+                    score_arg_sort = np.argsort(y_score)
+                    y_score_sorted = y_score[score_arg_sort]
+                    y_true_sorted = y_true[score_arg_sort]
+                    y_true_sorted_cumsum = np.cumsum(y_true_sorted)
+
+                    # unique thresholds
+                    (thresholds, unique_indices) = np.unique(
+                        y_score_sorted, return_index=True)
+                    num_prec_recall = len(unique_indices) + 1
+
+                    # prepare precision recall
+                    num_examples = len(y_score_sorted)
+                    # follow https://github.com/ScanNet/ScanNet/pull/26 ? # noqa
+                    num_true_examples = y_true_sorted_cumsum[-1] if len(
+                        y_true_sorted_cumsum) > 0 else 0
+                    precision = np.zeros(num_prec_recall)
+                    recall = np.zeros(num_prec_recall)
+
+                    # deal with the first point
+                    y_true_sorted_cumsum = np.append(y_true_sorted_cumsum, 0)
+                    # deal with remaining
+                    for idx_res, idx_scores in enumerate(unique_indices):
+                        cumsum = y_true_sorted_cumsum[idx_scores - 1]
+                        tp = num_true_examples - cumsum
+                        fp = num_examples - idx_scores - tp
+                        fn = cumsum + hard_false_negatives
+                        p = float(tp) / (tp + fp)
+                        r = float(tp) / (tp + fn)
+                        precision[idx_res] = p
+                        recall[idx_res] = r
+
+                    # first point in curve is artificial
+                    precision[-1] = 1.
+                    recall[-1] = 0.
+
+                    # compute average of precision-recall curve
+                    recall_for_conv = np.copy(recall)
+                    recall_for_conv = np.append(recall_for_conv[0],
+                                                recall_for_conv)
+                    recall_for_conv = np.append(recall_for_conv, 0.)
+
+                    stepWidths = np.convolve(recall_for_conv, [-0.5, 0, 0.5],
+                                             'valid')
+                    # integrate is now simply a dot product
+                    ap_current = np.dot(precision, stepWidths)
+
+                elif has_gt:
+                    ap_current = 0.0
+                else:
+                    ap_current = float('nan')
+                ap[di, li, oi] = ap_current
+    return ap
+
+
+def compute_averages(aps, options, class_labels):
+    """Averages AP scores for all categories.
+
+    Args:
+        aps (np.array): AP scores for all thresholds and categories.
+        options (dict): ScanNet evaluator options. See get_options.
+        class_labels (tuple[str]): Class names.
+
+    Returns:
+        dict: Overall and per-category AP scores.
+    """
+    d_inf = 0
+    o50 = np.where(np.isclose(options['overlaps'], 0.5))
+    o25 = np.where(np.isclose(options['overlaps'], 0.25))
+    o_all_but25 = np.where(
+        np.logical_not(np.isclose(options['overlaps'], 0.25)))
+    avg_dict = {}
+    avg_dict['all_ap'] = np.nanmean(aps[d_inf, :, o_all_but25])
+    avg_dict['all_ap_50%'] = np.nanmean(aps[d_inf, :, o50])
+    avg_dict['all_ap_25%'] = np.nanmean(aps[d_inf, :, o25])
+    avg_dict['classes'] = {}
+    for (li, label_name) in enumerate(class_labels):
+        avg_dict['classes'][label_name] = {}
+        avg_dict['classes'][label_name]['ap'] = np.average(aps[d_inf, li,
+                                                               o_all_but25])
+        avg_dict['classes'][label_name]['ap50%'] = np.average(aps[d_inf, li,
+                                                                  o50])
+        avg_dict['classes'][label_name]['ap25%'] = np.average(aps[d_inf, li,
+                                                                  o25])
+    return avg_dict
+
+
+def assign_instances_for_scan(pred_info, gt_ids, options, valid_class_ids,
+                              class_labels, id_to_label):
+    """Assign gt and predicted instances for a single scene.
+
+    Args:
+        pred_info (dict): Predicted masks, labels and scores.
+        gt_ids (np.array): Ground truth instance masks.
+        options (dict): ScanNet evaluator options. See get_options.
+        valid_class_ids (tuple[int]): Ids of valid categories.
+        class_labels (tuple[str]): Class names.
+        id_to_label (dict[int, str]): Mapping of valid class id to class label.
+
+    Returns:
+        dict: Per class assigned gt to predicted instances.
+        dict: Per class assigned predicted to gt instances.
+    """
+    # get gt instances
+    gt_instances = util_3d.get_instances(gt_ids, valid_class_ids, class_labels,
+                                         id_to_label)
+    # associate
+    gt2pred = deepcopy(gt_instances)
+    for label in gt2pred:
+        for gt in gt2pred[label]:
+            gt['matched_pred'] = []
+    pred2gt = {}
+    for label in class_labels:
+        pred2gt[label] = []
+    num_pred_instances = 0
+    # mask of void labels in the ground truth
+    bool_void = np.logical_not(np.in1d(gt_ids // 1000, valid_class_ids))
+    # go through all prediction masks
+    for pred_mask_file in pred_info:
+        label_id = int(pred_info[pred_mask_file]['label_id'])
+        conf = pred_info[pred_mask_file]['conf']
+        if not label_id in id_to_label:  # noqa E713
+            continue
+        label_name = id_to_label[label_id]
+        # read the mask
+        pred_mask = pred_info[pred_mask_file]['mask']
+        if len(pred_mask) != len(gt_ids):
+            raise ValueError('len(pred_mask) != len(gt_ids)')
+        # convert to binary
+        pred_mask = np.not_equal(pred_mask, 0)
+        num = np.count_nonzero(pred_mask)
+        if num < options['min_region_sizes'][0]:
+            continue  # skip if empty
+
+        pred_instance = {}
+        pred_instance['filename'] = pred_mask_file
+        pred_instance['pred_id'] = num_pred_instances
+        pred_instance['label_id'] = label_id
+        pred_instance['vert_count'] = num
+        pred_instance['confidence'] = conf
+        pred_instance['void_intersection'] = np.count_nonzero(
+            np.logical_and(bool_void, pred_mask))
+
+        # matched gt instances
+        matched_gt = []
+        # go through all gt instances with matching label
+        for (gt_num, gt_inst) in enumerate(gt2pred[label_name]):
+            intersection = np.count_nonzero(
+                np.logical_and(gt_ids == gt_inst['instance_id'], pred_mask))
+            if intersection > 0:
+                gt_copy = gt_inst.copy()
+                pred_copy = pred_instance.copy()
+                gt_copy['intersection'] = intersection
+                pred_copy['intersection'] = intersection
+                matched_gt.append(gt_copy)
+                gt2pred[label_name][gt_num]['matched_pred'].append(pred_copy)
+        pred_instance['matched_gt'] = matched_gt
+        num_pred_instances += 1
+        pred2gt[label_name].append(pred_instance)
+
+    return gt2pred, pred2gt
+
+
+def scannet_eval(preds, gts, options, valid_class_ids, class_labels,
+                 id_to_label):
+    """Evaluate instance segmentation in ScanNet protocol.
+
+    Args:
+        preds (list[dict]): Per scene predictions of mask, label and
+            confidence.
+        gts (list[np.array]): Per scene ground truth instance masks.
+        options (dict): ScanNet evaluator options. See get_options.
+        valid_class_ids (tuple[int]): Ids of valid categories.
+        class_labels (tuple[str]): Class names.
+        id_to_label (dict[int, str]): Mapping of valid class id to class label.
+
+    Returns:
+        dict: Overall and per-category AP scores.
+    """
+    options = get_options(options)
+    matches = {}
+    for i, (pred, gt) in enumerate(zip(preds, gts)):
+        matches_key = i
+        # assign gt to predictions
+        gt2pred, pred2gt = assign_instances_for_scan(pred, gt, options,
+                                                     valid_class_ids,
+                                                     class_labels, id_to_label)
+        matches[matches_key] = {}
+        matches[matches_key]['gt'] = gt2pred
+        matches[matches_key]['pred'] = pred2gt
+
+    ap_scores = evaluate_matches(matches, class_labels, options)
+    avgs = compute_averages(ap_scores, options, class_labels)
+    return avgs
+
+
+def get_options(options=None):
+    """Set ScanNet evaluator options.
+
+    Args:
+        options (dict, optional): Not default options. Default: None.
+
+    Returns:
+        dict: Updated options with all 4 keys.
+    """
+    assert options is None or isinstance(options, dict)
+    _options = dict(
+        overlaps=np.append(np.arange(0.5, 0.95, 0.05), 0.25),
+        min_region_sizes=np.array([100]),
+        distance_threshes=np.array([float('inf')]),
+        distance_confs=np.array([-float('inf')]))
+    if options is not None:
+        _options.update(options)
+    return _options
--- a/mmdet3d/core/evaluation/scannet_utils/util_3d.py
+++ b/mmdet3d/core/evaluation/scannet_utils/util_3d.py
+# Copyright (c) OpenMMLab. All rights reserved.
+# adapted from https://github.com/ScanNet/ScanNet/blob/master/BenchmarkScripts/util_3d.py # noqa
+import json
+
+import numpy as np
+
+
+class Instance:
+    """Single instance for ScanNet evaluator.
+
+    Args:
+        mesh_vert_instances (np.array): Instance ids for each point.
+        instance_id: Id of single instance.
+    """
+    instance_id = 0
+    label_id = 0
+    vert_count = 0
+    med_dist = -1
+    dist_conf = 0.0
+
+    def __init__(self, mesh_vert_instances, instance_id):
+        if instance_id == -1:
+            return
+        self.instance_id = int(instance_id)
+        self.label_id = int(self.get_label_id(instance_id))
+        self.vert_count = int(
+            self.get_instance_verts(mesh_vert_instances, instance_id))
+
+    @staticmethod
+    def get_label_id(instance_id):
+        return int(instance_id // 1000)
+
+    @staticmethod
+    def get_instance_verts(mesh_vert_instances, instance_id):
+        return (mesh_vert_instances == instance_id).sum()
+
+    def to_json(self):
+        return json.dumps(
+            self, default=lambda o: o.__dict__, sort_keys=True, indent=4)
+
+    def to_dict(self):
+        dict = {}
+        dict['instance_id'] = self.instance_id
+        dict['label_id'] = self.label_id
+        dict['vert_count'] = self.vert_count
+        dict['med_dist'] = self.med_dist
+        dict['dist_conf'] = self.dist_conf
+        return dict
+
+    def from_json(self, data):
+        self.instance_id = int(data['instance_id'])
+        self.label_id = int(data['label_id'])
+        self.vert_count = int(data['vert_count'])
+        if 'med_dist' in data:
+            self.med_dist = float(data['med_dist'])
+            self.dist_conf = float(data['dist_conf'])
+
+    def __str__(self):
+        return '(' + str(self.instance_id) + ')'
+
+
+def get_instances(ids, class_ids, class_labels, id2label):
+    """Transform gt instance mask to Instance objects.
+
+    Args:
+        ids (np.array): Instance ids for each point.
+        class_ids: (tuple[int]): Ids of valid categories.
+        class_labels (tuple[str]): Class names.
+        id2label: (dict[int, str]): Mapping of valid class id to class label.
+
+    Returns:
+        dict [str, list]: Instance objects grouped by class label.
+    """
+    instances = {}
+    for label in class_labels:
+        instances[label] = []
+    instance_ids = np.unique(ids)
+    for id in instance_ids:
+        if id == 0:
+            continue
+        inst = Instance(ids, id)
+        if inst.label_id in class_ids:
+            instances[id2label[inst.label_id]].append(inst.to_dict())
+    return instances
--- a/mmdet3d/datasets/__init__.py
+++ b/mmdet3d/datasets/__init__.py
@@ -21,7 +21,8 @@ from .pipelines import (AffineResize, BackgroundPointsFilter, GlobalAlignment,
                        VoxelBasedPointSampler)
 # yapf: enable
 from .s3dis_dataset import S3DISDataset, S3DISSegDataset
-from .scannet_dataset import ScanNetDataset, ScanNetSegDataset
+from .scannet_dataset import (ScanNetDataset, ScanNetInstanceSegDataset,
+                              ScanNetSegDataset)
 from .semantickitti_dataset import SemanticKITTIDataset
 from .sunrgbd_dataset import SUNRGBDDataset
 from .utils import get_loading_pipeline
@@ -35,10 +36,10 @@ __all__ = [
    'LoadPointsFromFile', 'S3DISSegDataset', 'S3DISDataset',
    'NormalizePointsColor', 'IndoorPatchPointSample', 'IndoorPointSample',
    'PointSample', 'LoadAnnotations3D', 'GlobalAlignment', 'SUNRGBDDataset',
-    'ScanNetDataset', 'ScanNetSegDataset', 'SemanticKITTIDataset',
-    'Custom3DDataset', 'Custom3DSegDataset', 'LoadPointsFromMultiSweeps',
-    'WaymoDataset', 'BackgroundPointsFilter', 'VoxelBasedPointSampler',
-    'get_loading_pipeline', 'RandomDropPointsColor', 'RandomJitterPoints',
-    'ObjectNameFilter', 'AffineResize', 'RandomShiftScale',
-    'LoadPointsFromDict'
+    'ScanNetDataset', 'ScanNetSegDataset', 'ScanNetInstanceSegDataset',
+    'SemanticKITTIDataset', 'Custom3DDataset', 'Custom3DSegDataset',
+    'LoadPointsFromMultiSweeps', 'WaymoDataset', 'BackgroundPointsFilter',
+    'VoxelBasedPointSampler', 'get_loading_pipeline', 'RandomDropPointsColor',
+    'RandomJitterPoints', 'ObjectNameFilter', 'AffineResize',
+    'RandomShiftScale', 'LoadPointsFromDict'
 ]
--- a/mmdet3d/datasets/scannet_dataset.py
+++ b/mmdet3d/datasets/scannet_dataset.py
@@ -5,7 +5,7 @@ from os import path as osp

 import numpy as np

-from mmdet3d.core import show_result, show_seg_result
+from mmdet3d.core import instance_seg_eval, show_result, show_seg_result
 from mmdet3d.core.bbox import DepthInstance3DBoxes
 from mmdet.datasets import DATASETS
 from mmseg.datasets import DATASETS as SEG_DATASETS
@@ -460,3 +460,151 @@ class ScanNetSegDataset(Custom3DSegDataset):
            outputs.append(dict(seg_mask=pred_label))

        return outputs, tmp_dir
+
+
+@DATASETS.register_module()
+@SEG_DATASETS.register_module()
+class ScanNetInstanceSegDataset(Custom3DSegDataset):
+    CLASSES = ('cabinet', 'bed', 'chair', 'sofa', 'table', 'door', 'window',
+               'bookshelf', 'picture', 'counter', 'desk', 'curtain',
+               'refrigerator', 'showercurtrain', 'toilet', 'sink', 'bathtub',
+               'garbagebin')
+
+    VALID_CLASS_IDS = (3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 14, 16, 24, 28, 33, 34,
+                       36, 39)
+
+    ALL_CLASS_IDS = tuple(range(41))
+
+    def get_ann_info(self, index):
+        """Get annotation info according to the given index.
+
+        Args:
+            index (int): Index of the annotation data to get.
+
+        Returns:
+            dict: annotation information consists of the following keys:
+                - pts_semantic_mask_path (str): Path of semantic masks.
+                - pts_instance_mask_path (str): Path of instance masks.
+        """
+        # Use index to get the annos, thus the evalhook could also use this api
+        info = self.data_infos[index]
+
+        pts_instance_mask_path = osp.join(self.data_root,
+                                          info['pts_instance_mask_path'])
+        pts_semantic_mask_path = osp.join(self.data_root,
+                                          info['pts_semantic_mask_path'])
+
+        anns_results = dict(
+            pts_instance_mask_path=pts_instance_mask_path,
+            pts_semantic_mask_path=pts_semantic_mask_path)
+        return anns_results
+
+    def get_classes_and_palette(self, classes=None, palette=None):
+        """Get class names of current dataset. Palette is simply ignored for
+        instance segmentation.
+
+        Args:
+            classes (Sequence[str] | str | None): If classes is None, use
+                default CLASSES defined by builtin dataset. If classes is a
+                string, take it as a file name. The file contains the name of
+                classes where each line contains one class name. If classes is
+                a tuple or list, override the CLASSES defined by the dataset.
+                Defaults to None.
+            palette (Sequence[Sequence[int]]] | np.ndarray | None):
+                The palette of segmentation map. If None is given, random
+                palette will be generated. Defaults to None.
+        """
+        if classes is not None:
+            return classes, None
+        return self.CLASSES, None
+
+    def _build_default_pipeline(self):
+        """Build the default pipeline for this dataset."""
+        pipeline = [
+            dict(
+                type='LoadPointsFromFile',
+                coord_type='DEPTH',
+                shift_height=False,
+                use_color=True,
+                load_dim=6,
+                use_dim=[0, 1, 2, 3, 4, 5]),
+            dict(
+                type='LoadAnnotations3D',
+                with_bbox_3d=False,
+                with_label_3d=False,
+                with_mask_3d=True,
+                with_seg_3d=True),
+            dict(
+                type='PointSegClassMapping',
+                valid_cat_ids=self.VALID_CLASS_IDS,
+                max_cat_id=40),
+            dict(
+                type='DefaultFormatBundle3D',
+                with_label=False,
+                class_names=self.CLASSES),
+            dict(
+                type='Collect3D',
+                keys=['points', 'pts_semantic_mask', 'pts_instance_mask'])
+        ]
+        return Compose(pipeline)
+
+    def evaluate(self,
+                 results,
+                 metric=None,
+                 options=None,
+                 logger=None,
+                 show=False,
+                 out_dir=None,
+                 pipeline=None):
+        """Evaluation in instance segmentation protocol.
+
+        Args:
+            results (list[dict]): List of results.
+            metric (str | list[str]): Metrics to be evaluated.
+            options (dict, optional): options for instance_seg_eval.
+            logger (logging.Logger | None | str): Logger used for printing
+                related information during evaluation. Defaults to None.
+            show (bool, optional): Whether to visualize.
+                Defaults to False.
+            out_dir (str, optional): Path to save the visualization results.
+                Defaults to None.
+            pipeline (list[dict], optional): raw data loading for showing.
+                Default: None.
+
+        Returns:
+            dict: Evaluation results.
+        """
+        assert isinstance(
+            results, list), f'Expect results to be list, got {type(results)}.'
+        assert len(results) > 0, 'Expect length of results > 0.'
+        assert len(results) == len(self.data_infos)
+        assert isinstance(
+            results[0], dict
+        ), f'Expect elements in results to be dict, got {type(results[0])}.'
+
+        load_pipeline = self._get_pipeline(pipeline)
+        pred_instance_masks = [result['instance_mask'] for result in results]
+        pred_instance_labels = [result['instance_label'] for result in results]
+        pred_instance_scores = [result['instance_score'] for result in results]
+        gt_semantic_masks, gt_instance_masks = zip(*[
+            self._extract_data(
+                index=i,
+                pipeline=load_pipeline,
+                key=['pts_semantic_mask', 'pts_instance_mask'],
+                load_annos=True) for i in range(len(self.data_infos))
+        ])
+        ret_dict = instance_seg_eval(
+            gt_semantic_masks,
+            gt_instance_masks,
+            pred_instance_masks,
+            pred_instance_labels,
+            pred_instance_scores,
+            valid_class_ids=self.VALID_CLASS_IDS,
+            class_labels=self.CLASSES,
+            options=options,
+            logger=logger)
+
+        if show:
+            raise NotImplementedError('show is not implemented for now')
+
+        return ret_dict
--- a/tests/test_data/test_datasets/test_scannet_dataset.py
+++ b/tests/test_data/test_datasets/test_scannet_dataset.py
@@ -5,7 +5,8 @@ import numpy as np
 import pytest
 import torch

-from mmdet3d.datasets import ScanNetDataset, ScanNetSegDataset
+from mmdet3d.datasets import (ScanNetDataset, ScanNetInstanceSegDataset,
+                              ScanNetSegDataset)


 def test_getitem():
@@ -685,3 +686,152 @@ def test_seg_format_results():
    expected_txt_path = osp.join(tmp_dir.name, 'results', 'scene0000_00.txt')
    assert np.all(result_files[0]['seg_mask'] == expected_label)
    mmcv.check_file_exist(expected_txt_path)
+
+
+def test_instance_seg_getitem():
+    np.random.seed(0)
+    root_path = './tests/data/scannet/'
+    ann_file = './tests/data/scannet/scannet_infos.pkl'
+    class_names = ('cabinet', 'bed', 'chair', 'sofa', 'table', 'door',
+                   'window', 'bookshelf', 'picture', 'counter', 'desk',
+                   'curtain', 'refrigerator', 'showercurtrain', 'toilet',
+                   'sink', 'bathtub', 'garbagebin')
+    train_pipeline = [
+        dict(
+            type='LoadPointsFromFile',
+            coord_type='DEPTH',
+            shift_height=False,
+            use_color=True,
+            load_dim=6,
+            use_dim=[0, 1, 2, 3, 4, 5]),
+        dict(
+            type='LoadAnnotations3D',
+            with_bbox_3d=False,
+            with_label_3d=False,
+            with_mask_3d=True,
+            with_seg_3d=True),
+        dict(
+            type='PointSegClassMapping',
+            valid_cat_ids=(3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 14, 16, 24, 28, 33,
+                           34, 36, 39),
+            max_cat_id=40),
+        dict(type='NormalizePointsColor', color_mean=None),
+        dict(type='DefaultFormatBundle3D', class_names=class_names),
+        dict(
+            type='Collect3D',
+            keys=['points', 'pts_semantic_mask', 'pts_instance_mask'])
+    ]
+    scannet_dataset = ScanNetInstanceSegDataset(
+        data_root=root_path,
+        ann_file=ann_file,
+        pipeline=train_pipeline,
+        classes=class_names,
+        test_mode=False)
+    expected_points = torch.tensor([[
+        -3.4742e+00, 7.8792e-01, 1.7397e+00, 3.3725e-01, 3.5294e-01, 3.0588e-01
+    ], [
+        2.7216e+00, 3.4164e+00, 2.4572e+00, 6.6275e-01, 6.2745e-01, 5.1373e-01
+    ],
+                                    [
+                                        1.3404e+00, -1.4675e+00, -4.4059e-02,
+                                        3.8431e-01, 3.6078e-01, 3.5686e-01
+                                    ],
+                                    [
+                                        -3.0335e+00, 2.7273e+00, 1.5181e+00,
+                                        2.3137e-01, 1.6078e-01, 8.2353e-02
+                                    ],
+                                    [
+                                        -4.3207e-01, 1.8154e+00, 1.7455e-01,
+                                        4.0392e-01, 3.8039e-01, 4.1961e-01
+                                    ]])
+    expected_semantic_mask = torch.tensor([11, 18, 18, 0, 4]).long()
+    expected_instance_mask = torch.tensor([6, 56, 10, 9, 35]).long()
+
+    data = scannet_dataset[0]
+    assert torch.allclose(data['points']._data[:5], expected_points, 1e-2)
+    assert torch.allclose(data['pts_semantic_mask']._data[:5],
+                          expected_semantic_mask)
+    assert torch.allclose(data['pts_instance_mask']._data[:5],
+                          expected_instance_mask)
+
+
+def test_instance_seg_evaluate():
+    root_path = './tests/data/scannet'
+    ann_file = './tests/data/scannet/scannet_infos.pkl'
+    class_names = ('cabinet', 'bed', 'chair', 'sofa', 'table', 'door',
+                   'window', 'bookshelf', 'picture', 'counter', 'desk',
+                   'curtain', 'refrigerator', 'showercurtrain', 'toilet',
+                   'sink', 'bathtub', 'garbagebin')
+    test_pipeline = [
+        dict(
+            type='LoadPointsFromFile',
+            coord_type='DEPTH',
+            shift_height=False,
+            use_color=True,
+            load_dim=6,
+            use_dim=[0, 1, 2, 3, 4, 5]),
+        dict(type='NormalizePointsColor', color_mean=None),
+        dict(type='DefaultFormatBundle3D', class_names=class_names),
+        dict(type='Collect3D', keys=['points'])
+    ]
+    scannet_dataset = ScanNetInstanceSegDataset(
+        data_root=root_path,
+        ann_file=ann_file,
+        pipeline=test_pipeline,
+        test_mode=True)
+
+    pred_mask = torch.tensor([
+        1, -1, -1, -1, 7, 11, 2, -1, 1, 10, -1, -1, 5, -1, -1, -1, -1, 1, -1,
+        -1, -1, -1, 0, -1, 1, -1, 12, -1, -1, -1, 8, 5, 1, 5, 2, -1, -1, -1,
+        -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 1, 8, -1, -1, -1,
+        0, 4, 3, -1, 9, -1, -1, 6, -1, -1, -1, -1, 13, -1, -1, 5, -1, 5, -1,
+        -1, 9, 0, 5, -1, -1, 2, 3, 4, -1, -1, -1, 2, -1, -1, -1, 5, 9, -1, 1,
+        -1, 4, 10, 4, -1
+    ]).long()
+    pred_labels = torch.tensor(
+        [4, 11, 11, 10, 0, 3, 12, 4, 14, 1, 0, 0, 0, 5, 5]).long()
+    pred_scores = torch.tensor([.99 for _ in range(len(pred_labels))])
+    results = [
+        dict(
+            instance_mask=pred_mask,
+            instance_label=pred_labels,
+            instance_score=torch.tensor(pred_scores))
+    ]
+    eval_pipeline = [
+        dict(
+            type='LoadPointsFromFile',
+            coord_type='DEPTH',
+            shift_height=False,
+            use_color=True,
+            load_dim=6,
+            use_dim=[0, 1, 2, 3, 4, 5]),
+        dict(
+            type='LoadAnnotations3D',
+            with_bbox_3d=False,
+            with_label_3d=False,
+            with_mask_3d=True,
+            with_seg_3d=True),
+        dict(
+            type='PointSegClassMapping',
+            valid_cat_ids=(3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 14, 16, 24, 28, 33,
+                           34, 36, 39),
+            max_cat_id=40),
+        dict(type='NormalizePointsColor', color_mean=None),
+        dict(type='DefaultFormatBundle3D', class_names=class_names),
+        dict(
+            type='Collect3D',
+            keys=['points', 'pts_semantic_mask', 'pts_instance_mask'])
+    ]
+    # We add options here as default min_region_size
+    # is much bigger than test instances.
+    ret_dict = scannet_dataset.evaluate(
+        results,
+        pipeline=eval_pipeline,
+        options=dict(min_region_sizes=np.array([1])))
+    assert abs(ret_dict['all_ap'] - 0.90625) < 0.001
+    assert abs(ret_dict['all_ap_50%'] - 0.90625) < 0.001
+    assert abs(ret_dict['all_ap_25%'] - 0.94444) < 0.001
+    assert abs(ret_dict['classes']['cabinet']['ap25%'] - 1.0) < 0.001
+    assert abs(ret_dict['classes']['cabinet']['ap50%'] - 0.65625) < 0.001
+    assert abs(ret_dict['classes']['door']['ap25%'] - 0.5) < 0.001
+    assert abs(ret_dict['classes']['door']['ap50%'] - 0.5) < 0.001
--- a/tests/test_metrics/test_instance_seg_eval.py
+++ b/tests/test_metrics/test_instance_seg_eval.py
+# Copyright (c) OpenMMLab. All rights reserved.
+import numpy as np
+import torch
+
+from mmdet3d.core import instance_seg_eval
+
+
+def test_instance_seg_eval():
+    valid_class_ids = (3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 14, 16, 24, 28, 33, 34,
+                       36, 39)
+    class_labels = ('cabinet', 'bed', 'chair', 'sofa', 'table', 'door',
+                    'window', 'bookshelf', 'picture', 'counter', 'desk',
+                    'curtain', 'refrigerator', 'showercurtrain', 'toilet',
+                    'sink', 'bathtub', 'garbagebin')
+    n_points_list = [3300, 3000]
+    gt_labels_list = [[0, 0, 0, 0, 0, 0, 14, 14, 2, 1],
+                      [13, 13, 2, 1, 3, 3, 0, 0, 0]]
+    gt_instance_masks = []
+    gt_semantic_masks = []
+    pred_instance_masks = []
+    pred_instance_labels = []
+    pred_instance_scores = []
+    for n_points, gt_labels in zip(n_points_list, gt_labels_list):
+        gt_instance_mask = np.ones(n_points, dtype=np.int) * -1
+        gt_semantic_mask = np.ones(n_points, dtype=np.int) * -1
+        pred_instance_mask = np.ones(n_points, dtype=np.int) * -1
+        labels = []
+        scores = []
+        for i, gt_label in enumerate(gt_labels):
+            begin = i * 300
+            end = begin + 300
+            gt_instance_mask[begin:end] = i
+            gt_semantic_mask[begin:end] = gt_label
+            pred_instance_mask[begin:end] = i
+            labels.append(gt_label)
+            scores.append(.99)
+        gt_instance_masks.append(torch.tensor(gt_instance_mask))
+        gt_semantic_masks.append(torch.tensor(gt_semantic_mask))
+        pred_instance_masks.append(torch.tensor(pred_instance_mask))
+        pred_instance_labels.append(torch.tensor(labels))
+        pred_instance_scores.append(torch.tensor(scores))
+
+    ret_value = instance_seg_eval(
+        gt_semantic_masks=gt_semantic_masks,
+        gt_instance_masks=gt_instance_masks,
+        pred_instance_masks=pred_instance_masks,
+        pred_instance_labels=pred_instance_labels,
+        pred_instance_scores=pred_instance_scores,
+        valid_class_ids=valid_class_ids,
+        class_labels=class_labels)
+    for label in [
+            'cabinet', 'bed', 'chair', 'sofa', 'showercurtrain', 'toilet'
+    ]:
+        metrics = ret_value['classes'][label]
+        assert metrics['ap'] == 1.0
+        assert metrics['ap50%'] == 1.0
+        assert metrics['ap25%'] == 1.0
+
+    pred_instance_masks[1][2240:2700] = -1
+    pred_instance_masks[0][2700:3000] = 8
+    pred_instance_labels[0][9] = 2
+    ret_value = instance_seg_eval(
+        gt_semantic_masks=gt_semantic_masks,
+        gt_instance_masks=gt_instance_masks,
+        pred_instance_masks=pred_instance_masks,
+        pred_instance_labels=pred_instance_labels,
+        pred_instance_scores=pred_instance_scores,
+        valid_class_ids=valid_class_ids,
+        class_labels=class_labels)
+    assert abs(ret_value['classes']['cabinet']['ap50%'] - 0.72916) < 0.01
+    assert abs(ret_value['classes']['cabinet']['ap25%'] - 0.88888) < 0.01
+    assert abs(ret_value['classes']['bed']['ap50%'] - 0.5) < 0.01
+    assert abs(ret_value['classes']['bed']['ap25%'] - 0.5) < 0.01
+    assert abs(ret_value['classes']['chair']['ap50%'] - 0.375) < 0.01
+    assert abs(ret_value['classes']['chair']['ap25%'] - 1.0) < 0.01