release groupfree3d models on scannet (#710)

2f95eb70 · hjin2902 · GitHub · f5cdc7b9 · 2f95eb70 · 2f95eb70
Unverified Commit 2f95eb70 authored Jul 12, 2021 by hjin2902 Committed by GitHub Jul 12, 2021
5 changed files
--- a/.dev_scripts/gather_models.py
+++ b/.dev_scripts/gather_models.py
@@ -25,7 +25,8 @@ SCHEDULES_LUT = {
    '_50e_': 50,
    '_80e_': 80,
    '_200e_': 200,
-    '_250e_': 250
+    '_250e_': 250,
+    '_400e_': 400
 }

 # TODO: add support for lyft dataset
@@ -40,8 +41,8 @@ RESULTS_LUT = {
    'lyft': ['score'],
    'scannet_seg': ['miou'],
    's3dis_seg': ['miou'],
-    'scannet': ['mAR_0.50'],
-    'sunrgbd': ['mAR_0.50']
+    'scannet': ['mAP_0.50'],
+    'sunrgbd': ['mAP_0.50']
 }



--- a/README.md
+++ b/README.md
@@ -91,6 +91,7 @@ Support methods
 - [x] [ImVoteNet (CVPR'2020)](configs/imvotenet/README.md)
 - [x] [FCOS3D (Arxiv'2021)](configs/fcos3d/README.md)
 - [x] [PointNet++ (NeurIPS'2017)](configs/pointnet2/README.md)
+- [x] [Group-Free-3D (Arxiv'2021)](configs/groupfree3d/README.md)

 |                    | ResNet   | ResNeXt  | SENet    |PointNet++ | HRNet | RegNetX | Res2Net |
 |--------------------|:--------:|:--------:|:--------:|:---------:|:-----:|:--------:|:-----:|
@@ -107,6 +108,7 @@ Support methods
 | ImVoteNet            | ✗        | ✗        | ✗        | ✓         | ✗     | ✗        | ✗     |
 | FCOS3D               | ✓        | ☐        | ☐        | ✗         | ☐     | ☐        | ☐     |
 | PointNet++           | ✗        | ✗        | ✗        | ✓         | ✗     | ✗        | ✗     |
+| Group-Free-3D        | ✗        | ✗        | ✗        | ✓         | ✗     | ✗        | ✗     |

 Other features
 - [x] [Dynamic Voxelization](configs/dynamic_voxelization/README.md)

--- a/configs/groupfree3d/README.md
+++ b/configs/groupfree3d/README.md
+# Group-Free 3D Object Detection via Transformers
+
+## Introduction
+
+<!-- [ALGORITHM] -->
+
+We implement Group-Free-3D and provide the result and checkpoints on ScanNet datasets.
+
+```
+@article{liu2021,
+  title={Group-Free 3D Object Detection via Transformers},
+  author={Liu, Ze and Zhang, Zheng and Cao, Yue and Hu, Han and Tong, Xin},
+  journal={arXiv preprint arXiv:2104.00678},
+  year={2021}
+}
+```
+
+## Results
+
+### ScanNet
+
+|  Method  |  Backbone   | Lr schd | Mem (GB) | Inf time (fps) | AP@0.25 |AP@0.5| Download |
+| :------: | :---------: | :-----: | :------: | :------------: | :----: |:----: | :------: |
+| [L6, O256](./groupfree3d_8x4_scannet-3d-18class-L6-O256.py ) |    PointNet++     |  3x    |6.7||65.59 (65.67*)|48.43 (47.74*)|[model](https://download.openmmlab.com/mmdetection3d/v0.1.0_models/groupfree3d/groupfree3d_8x4_scannet-3d-18class-L6-O256/groupfree3d_8x4_scannet-3d-18class-L6-O256_20210702_145347-3499eb55.pth) &#124; [log](https://download.openmmlab.com/mmdetection3d/v0.1.0_models/groupfree3d/groupfree3d_8x4_scannet-3d-18class-L6-O256/groupfree3d_8x4_scannet-3d-18class-L6-O256_20210702_145347.log.json)|
+| [L12, O256](./groupfree3d_8x4_scannet-3d-18class-L12-O256.py ) |    PointNet++     |  3x    |9.4||67.68 (66.22*)|49.30 (48.95*)|[model](https://download.openmmlab.com/mmdetection3d/v0.1.0_models/groupfree3d/groupfree3d_8x4_scannet-3d-18class-L12-O256/groupfree3d_8x4_scannet-3d-18class-L12-O256_20210702_150907-1c5551ad.pth) &#124; [log](https://download.openmmlab.com/mmdetection3d/v0.1.0_models/groupfree3d/groupfree3d_8x4_scannet-3d-18class-L12-O256/groupfree3d_8x4_scannet-3d-18class-L12-O256_20210702_150907.log.json)|
+| [L12, O256](./groupfree3d_8x4_scannet-3d-18class-w2x-L12-O256.py ) |    PointNet++w2x     |  3x    |13.3||67.09 (67.30*)|50.76 (50.44*)|[model](https://download.openmmlab.com/mmdetection3d/v0.1.0_models/groupfree3d/groupfree3d_8x4_scannet-3d-18class-w2x-L12-O256/groupfree3d_8x4_scannet-3d-18class-w2x-L12-O256_20210702_200301-944f0ac0.pth) &#124; [log](https://download.openmmlab.com/mmdetection3d/v0.1.0_models/groupfree3d/groupfree3d_8x4_scannet-3d-18class-w2x-L12-O256/groupfree3d_8x4_scannet-3d-18class-w2x-L12-O256_20210702_200301.log.json)|
+| [L12, O512](./groupfree3d_8x4_scannet-3d-18class-w2x-L12-O512.py ) |    PointNet++w2x     |  3x    |18.8||68.31 (68.20*)|51.73 (51.31*)|[model](https://download.openmmlab.com/mmdetection3d/v0.1.0_models/groupfree3d/groupfree3d_8x4_scannet-3d-18class-w2x-L12-O512/groupfree3d_8x4_scannet-3d-18class-w2x-L12-O512_20210702_220204-187b71c7.pth) &#124; [log](https://download.openmmlab.com/mmdetection3d/v0.1.0_models/groupfree3d/groupfree3d_8x4_scannet-3d-18class-w2x-L12-O512/groupfree3d_8x4_scannet-3d-18class-w2x-L12-O512_20210702_220204.log.json)|
+
+**Notes:**
+
+- We report the best results (AP@0.50) on validation set during each training. * means the evaluation method in the paper: we train each setting 5 times and test each training trial 5 times, then the average performance of these 25 trials is reported to account for algorithm randomness.
+- We use 4 GPUs for training by default as the original code.
--- a/configs/groupfree3d/metafile.yml
+++ b/configs/groupfree3d/metafile.yml
+Collections:
+  - Name: Group-Free-3D
+    Metadata:
+      Training Techniques:
+        - AdamW
+      Training Resources: 4x V100 GPUs
+      Architecture:
+        - PointNet++
+    Paper: https://arxiv.org/abs/2104.00678
+    README: configs/groupfree3d/README.md
+
+Models:
+  - Name: groupfree3d_8x4_scannet-3d-18class-L6-O256.py
+    In Collection: Group-Free-3D
+    Config: configs/groupfree3d/groupfree3d_8x4_scannet-3d-18class-L6-O256.py
+    Metadata:
+      Training Data: ScanNet
+      Training Memory (GB): 6.7
+    Results:
+      - Task: 3D Object Detection
+        Dataset: ScanNet
+        Metrics:
+          AP@0.25: 65.59
+          AP@0.5: 48.43
+    Weights: https://download.openmmlab.com/mmdetection3d/v0.1.0_models/groupfree3d/groupfree3d_8x4_scannet-3d-18class-L6-O256/groupfree3d_8x4_scannet-3d-18class-L6-O256_20210702_145347-3499eb55.pth
+
+  - Name: groupfree3d_8x4_scannet-3d-18class-L12-O256.py
+    In Collection: Group-Free-3D
+    Config: configs/groupfree3d/groupfree3d_8x4_scannet-3d-18class-L12-O256.py
+    Metadata:
+      Training Data: ScanNet
+      Training Memory (GB): 9.4
+    Results:
+      - Task: 3D Object Detection
+        Dataset: ScanNet
+        Metrics:
+          AP@0.25: 67.68
+          AP@0.5: 49.30
+    Weights: https://download.openmmlab.com/mmdetection3d/v0.1.0_models/groupfree3d/groupfree3d_8x4_scannet-3d-18class-L12-O256/groupfree3d_8x4_scannet-3d-18class-L12-O256_20210702_150907-1c5551ad.pth
+
+  - Name: groupfree3d_8x4_scannet-3d-18class-w2x-L12-O256.py
+    In Collection: Group-Free-3D
+    Config: configs/groupfree3d/groupfree3d_8x4_scannet-3d-18class-w2x-L12-O256.py
+    Metadata:
+      Training Data: ScanNet
+      Training Memory (GB): 13.3
+    Results:
+      - Task: 3D Object Detection
+        Dataset: ScanNet
+        Metrics:
+          AP@0.25: 67.09
+          AP@0.5: 50.76
+    Weights: https://download.openmmlab.com/mmdetection3d/v0.1.0_models/groupfree3d/groupfree3d_8x4_scannet-3d-18class-w2x-L12-O256/groupfree3d_8x4_scannet-3d-18class-w2x-L12-O256_20210702_200301-944f0ac0.pth
+
+  - Name: groupfree3d_8x4_scannet-3d-18class-w2x-L12-O512.py
+    In Collection: Group-Free-3D
+    Config: configs/groupfree3d/groupfree3d_8x4_scannet-3d-18class-w2x-L12-O512.py
+    Metadata:
+      Training Data: ScanNet
+      Training Memory (GB): 18.8
+    Results:
+      - Task: 3D Object Detection
+        Dataset: ScanNet
+        Metrics:
+          AP@0.25: 68.31
+          AP@0.5: 51.73
+    Weights: https://download.openmmlab.com/mmdetection3d/v0.1.0_models/groupfree3d/groupfree3d_8x4_scannet-3d-18class-w2x-L12-O512/groupfree3d_8x4_scannet-3d-18class-w2x-L12-O512_20210702_220204-187b71c7.pth
--- a/docs/model_zoo.md
+++ b/docs/model_zoo.md
@@ -65,3 +65,7 @@ Please refer to [FCOS3D](https://github.com/open-mmlab/mmdetection3d/blob/master
 ### PointNet++

 Please refer to [PointNet++](https://github.com/open-mmlab/mmdetection3d/blob/master/configs/pointnet2) for details. We provide PointNet++ baselines on ScanNet and S3DIS datasets.
+
+### Group-Free-3D
+
+Please refer to [Group-Free-3D](https://github.com/open-mmlab/mmdetection3d/blob/master/configs/groupfree3d) for details. We provide Group-Free-3D baselines on ScanNet datasets.