Merge branch 'master' into v1.0.0.dev0

docs_zh-CN/tutorials/config.md → docs/zh_cn/tutorials/config.md

@@ -215,13 +215,13 @@ train_pipeline = [  # 训练流水线，更多细节请参考 mmdet3d.datasets.p
         rot_range=[-0.027777777777777776, 0.027777777777777776],  # 旋转角范围
         scale_range=None),  # 缩放尺寸范围
     dict(
-        type='DefaultFormatBundle3D',  # 默认格式打包以收集读取的所有数据，更多细节请参考 mmdet3d.datasets.pipelines.formating
+        type='DefaultFormatBundle3D',  # 默认格式打包以收集读取的所有数据，更多细节请参考 mmdet3d.datasets.pipelines.formatting
         class_names=('cabinet', 'bed', 'chair', 'sofa', 'table', 'door',
                      'window', 'bookshelf', 'picture', 'counter', 'desk',
                      'curtain', 'refrigerator', 'showercurtrain', 'toilet',
                      'sink', 'bathtub', 'garbagebin')),
     dict(
-        type='Collect3D',  # 最后一个流程，决定哪些键值对应的数据会被输入给检测器，更多细节请参考 mmdet3d.datasets.pipelines.formating
+        type='Collect3D',  # 最后一个流程，决定哪些键值对应的数据会被输入给检测器，更多细节请参考 mmdet3d.datasets.pipelines.formatting
         keys=[
             'points', 'gt_bboxes_3d', 'gt_labels_3d', 'pts_semantic_mask',
             'pts_instance_mask'
@@ -236,12 +236,12 @@ test_pipeline = [  # 测试流水线，更多细节请参考 mmdet3d.datasets.pi
     dict(type='PointSample',  # 室内点采样，更多细节请参考 mmdet3d.datasets.pipelines.indoor_sample
             num_points=40000),  # 采样的点的数量
     dict(
-        type='DefaultFormatBundle3D',  # 默认格式打包以收集读取的所有数据，更多细节请参考 mmdet3d.datasets.pipelines.formating
+        type='DefaultFormatBundle3D',  # 默认格式打包以收集读取的所有数据，更多细节请参考 mmdet3d.datasets.pipelines.formatting
         class_names=('cabinet', 'bed', 'chair', 'sofa', 'table', 'door',
                      'window', 'bookshelf', 'picture', 'counter', 'desk',
                      'curtain', 'refrigerator', 'showercurtrain', 'toilet',
                      'sink', 'bathtub', 'garbagebin')),
-    dict(type='Collect3D',  # 最后一个流程，决定哪些键值对应的数据会被输入给检测器，更多细节请参考 mmdet3d.datasets.pipelines.formating
+    dict(type='Collect3D',  # 最后一个流程，决定哪些键值对应的数据会被输入给检测器，更多细节请参考 mmdet3d.datasets.pipelines.formatting
         keys=['points'])
 ]
 eval_pipeline = [  # 模型验证或可视化所使用的流水线，更多细节请参考 mmdet3d.datasets.pipelines
@@ -251,13 +251,13 @@ eval_pipeline = [  # 模型验证或可视化所使用的流水线，更多细
         load_dim=6,  # 读取的点的维度
         use_dim=[0, 1, 2]),  # 使用所读取点的哪些维度
     dict(
-        type='DefaultFormatBundle3D',  # 默认格式打包以收集读取的所有数据，更多细节请参考 mmdet3d.datasets.pipelines.formating
+        type='DefaultFormatBundle3D',  # 默认格式打包以收集读取的所有数据，更多细节请参考 mmdet3d.datasets.pipelines.formatting
         class_names=('cabinet', 'bed', 'chair', 'sofa', 'table', 'door',
                      'window', 'bookshelf', 'picture', 'counter', 'desk',
                      'curtain', 'refrigerator', 'showercurtrain', 'toilet',
                      'sink', 'bathtub', 'garbagebin')),
         with_label=False),
-    dict(type='Collect3D',  # 最后一个流程，决定哪些键值对应的数据会被输入给检测器，更多细节请参考 mmdet3d.datasets.pipelines.formating
+    dict(type='Collect3D',  # 最后一个流程，决定哪些键值对应的数据会被输入给检测器，更多细节请参考 mmdet3d.datasets.pipelines.formatting
         keys=['points'])
 ]
 data = dict(

mmdet3d/__init__.py

@@ -19,7 +19,7 @@ def digit_version(version_str):
 
 
 mmcv_minimum_version = '1.3.8'
-mmcv_maximum_version = '1.4.0'
+mmcv_maximum_version = '1.5.0'
 mmcv_version = digit_version(mmcv.__version__)
 
 

mmdet3d/apis/inference.py

@@ -65,6 +65,7 @@ def init_model(config, checkpoint=None, device='cuda:0'):
         if 'PALETTE' in checkpoint['meta']:  # 3D Segmentor
             model.PALETTE = checkpoint['meta']['PALETTE']
     model.cfg = config  # save the config in the model for convenience
+    torch.cuda.set_device(device)
     model.to(device)
     model.eval()
     return model

mmdet3d/apis/test.py

@@ -44,7 +44,12 @@ def single_gpu_test(model,
             models_3d = (Base3DDetector, Base3DSegmentor,
                          SingleStageMono3DDetector)
             if isinstance(model.module, models_3d):
-                model.module.show_results(data, result, out_dir=out_dir)
+                model.module.show_results(
+                    data,
+                    result,
+                    out_dir=out_dir,
+                    show=show,
+                    score_thr=show_score_thr)
             # Visualize the results of MMDetection model
             # 'show_result' is MMdetection visualization API
             else:

mmdet3d/core/visualizer/show_result.py

@@ -78,7 +78,8 @@ def show_result(points,
                 out_dir,
                 filename,
                 show=False,
-                snapshot=False):
+                snapshot=False,
+                pred_labels=None):
     """Convert results into format that is directly readable for meshlab.
 
     Args:
@@ -87,10 +88,11 @@ def show_result(points,
         pred_bboxes (np.ndarray): Predicted boxes.
         out_dir (str): Path of output directory
         filename (str): Filename of the current frame.
-        show (bool, optional): Visualize the results online.
-            Defaults to False.
+        show (bool, optional): Visualize the results online. Defaults to False.
         snapshot (bool, optional): Whether to save the online results.
             Defaults to False.
+        pred_labels (np.ndarray, optional): Predicted labels of boxes.
+            Defaults to None.
     """
     result_path = osp.join(out_dir, filename)
     mmcv.mkdir_or_exist(result_path)
@@ -100,7 +102,23 @@ def show_result(points,
 
         vis = Visualizer(points)
         if pred_bboxes is not None:
+            if pred_labels is None:
                 vis.add_bboxes(bbox3d=pred_bboxes)
+            else:
+                palette = np.random.randint(
+                    0, 255, size=(pred_labels.max() + 1, 3)) / 256
+                labelDict = {}
+                for j in range(len(pred_labels)):
+                    i = int(pred_labels[j].numpy())
+                    if labelDict.get(i) is None:
+                        labelDict[i] = []
+                    labelDict[i].append(pred_bboxes[j])
+                for i in labelDict:
+                    vis.add_bboxes(
+                        bbox3d=np.array(labelDict[i]),
+                        bbox_color=palette[i],
+                        points_in_box_color=palette[i])
+
         if gt_bboxes is not None:
             vis.add_bboxes(bbox3d=gt_bboxes, bbox_color=(0, 0, 1))
         show_path = osp.join(result_path,

mmdet3d/models/detectors/base.py

@@ -60,13 +60,18 @@ class Base3DDetector(BaseDetector):
         else:
             return self.forward_test(**kwargs)
 
-    def show_results(self, data, result, out_dir):
+    def show_results(self, data, result, out_dir, show=False, score_thr=None):
         """Results visualization.
 
         Args:
             data (list[dict]): Input points and the information of the sample.
             result (list[dict]): Prediction results.
             out_dir (str): Output directory of visualization result.
+            show (bool, optional): Determines whether you are
+                going to show result by open3d.
+                Defaults to False.
+            score_thr (float, optional): Score threshold of bounding boxes.
+                Default to None.
         """
         for batch_id in range(len(result)):
             if isinstance(data['points'][0], DC):
@@ -93,6 +98,12 @@ class Base3DDetector(BaseDetector):
             assert out_dir is not None, 'Expect out_dir, got none.'
 
             pred_bboxes = result[batch_id]['boxes_3d']
+            pred_labels = result[batch_id]['labels_3d']
+
+            if score_thr is not None:
+                mask = result[batch_id]['scores_3d'] > score_thr
+                pred_bboxes = pred_bboxes[mask]
+                pred_labels = pred_labels[mask]
 
             # for now we convert points and bbox into depth mode
             if (box_mode_3d == Box3DMode.CAM) or (box_mode_3d
@@ -105,4 +116,11 @@ class Base3DDetector(BaseDetector):
                 ValueError(
                     f'Unsupported box_mode_3d {box_mode_3d} for conversion!')
             pred_bboxes = pred_bboxes.tensor.cpu().numpy()
-            show_result(points, None, pred_bboxes, out_dir, file_name)
+            show_result(
+                points,
+                None,
+                pred_bboxes,
+                out_dir,
+                file_name,
+                show=show,
+                pred_labels=pred_labels)

mmdet3d/models/detectors/single_stage_mono3d.py

@@ -179,13 +179,20 @@ class SingleStageMono3DDetector(SingleStageDetector):
 
         return [bbox_list]
 
-    def show_results(self, data, result, out_dir):
+    def show_results(self, data, result, out_dir, show=False, score_thr=None):
         """Results visualization.
 
         Args:
             data (list[dict]): Input images and the information of the sample.
             result (list[dict]): Prediction results.
             out_dir (str): Output directory of visualization result.
+            show (bool, optional): Determines whether you are
+                going to show result by open3d.
+                Defaults to False.
+            TODO: implement score_thr of single_stage_mono3d.
+            score_thr (float, optional): Score threshold of bounding boxes.
+                Default to None.
+                Not implemented yet, but it is here for unification.
         """
         for batch_id in range(len(result)):
             if isinstance(data['img_metas'][0], DC):
@@ -216,4 +223,4 @@ class SingleStageMono3DDetector(SingleStageDetector):
                 out_dir,
                 file_name,
                 'camera',
-                show=True)
+                show=show)

mmdet3d/models/segmentors/base.py

@@ -72,7 +72,9 @@ class Base3DSegmentor(BaseSegmentor):
                      result,
                      palette=None,
                      out_dir=None,
-                     ignore_index=None):
+                     ignore_index=None,
+                     show=False,
+                     score_thr=None):
         """Results visualization.
 
         Args:
@@ -85,6 +87,13 @@ class Base3DSegmentor(BaseSegmentor):
             ignore_index (int, optional): The label index to be ignored, e.g.
                 unannotated points. If None is given, set to len(self.CLASSES).
                 Defaults to None.
+            show (bool, optional): Determines whether you are
+                going to show result by open3d.
+                Defaults to False.
+            TODO: implement score_thr of Base3DSegmentor.
+            score_thr (float, optional): Score threshold of bounding boxes.
+                Default to None.
+                Not implemented yet, but it is here for unification.
         """
         assert out_dir is not None, 'Expect out_dir, got none.'
         if palette is None:
@@ -123,4 +132,4 @@ class Base3DSegmentor(BaseSegmentor):
                 file_name,
                 palette,
                 ignore_index,
-                show=True)
+                show=show)

mmdet3d/version.py

 # Copyright (c) Open-MMLab. All rights reserved.
 
-__version__ = '0.17.2'
+__version__ = '0.17.3'
 short_version = __version__
 
 

requirements/docs.txt

 docutils==0.16.0
 m2r
-myst-parser
-opencv-python
+mistune==0.8.4
 -e git+https://github.com/open-mmlab/pytorch_sphinx_theme.git#egg=pytorch_sphinx_theme
+recommonmark
 sphinx==4.0.2
 sphinx-copybutton
 sphinx_markdown_tables
-torch

requirements/mminstall.txt

-mmcv-full>=1.3.8,<=1.4.0
+mmcv-full>=1.3.8,<=1.5.0
 mmdet>=2.14.0,<=3.0.0
 mmsegmentation>=0.14.1,<=1.0.0

requirements/readthedocs.txt

 mmcv
+mmdet
+mmsegmentation>=0.14.1
 torch
 torchvision