[Fix] fix some mono3d related bugs (#1816)

* fix mono3d related bugs

* update kitti-mono3d script

* update mono3d task

* update resize3d and randomresize3d

* fix

* update dataset converter script

* fix part of comments

* unify the task name in datasets and visualization

* fix comments

* rename 3d to lidar_det

* fix ci

* change boxlist to boxtype

* change default value ot lidar_det

* fix bugs

Author: ZCMax

configs/_base_/datasets/kitti-mono3d.py

@@ -35,6 +35,10 @@
     dict(type='Resize', scale=(1242, 375), keep_ratio=True),
     dict(type='Pack3DDetInputs', keys=['img'])
 ]
+eval_pipeline = [
+    dict(type='LoadImageFromFileMono3D'),
+    dict(type='Pack3DDetInputs', keys=['img'])
+]
 
 train_dataloader = dict(
     batch_size=2,

configs/_base_/datasets/nus-mono3d.py

@@ -65,7 +65,7 @@
             CAM_BACK_RIGHT='samples/CAM_BACK_RIGHT',
             CAM_BACK_LEFT='samples/CAM_BACK_LEFT'),
         ann_file='nuscenes_infos_train.pkl',
-        task='mono3d',
+        task='mono_det',
         pipeline=train_pipeline,
         metainfo=metainfo,
         modality=input_modality,
@@ -92,7 +92,7 @@
             CAM_BACK_RIGHT='samples/CAM_BACK_RIGHT',
             CAM_BACK_LEFT='samples/CAM_BACK_LEFT'),
         ann_file='nuscenes_infos_val.pkl',
-        task='mono3d',
+        task='mono_det',
         pipeline=test_pipeline,
         modality=input_modality,
         metainfo=metainfo,

configs/pgd/pgd_r101-caffe_fpn_head-gn_4xb3-4x_kitti-mono3d.py

@@ -113,7 +113,7 @@
 
 # optimizer
 optim_wrapper = dict(
-    optimizer=dict(lr=0.01),
+    optimizer=dict(lr=0.001),
     paramwise_cfg=dict(bias_lr_mult=2., bias_decay_mult=0.),
     clip_grad=dict(max_norm=35, norm_type=2))
 
@@ -134,4 +134,4 @@
         gamma=0.1)
 ]
 
-train_cfg = dict(max_epochs=48)
+train_cfg = dict(max_epochs=48, val_interval=2)

configs/smoke/smoke_dla34_dlaneck_gn-all_4xb8-6x_kitti-mono3d.py

@@ -47,8 +47,10 @@
 test_dataloader = dict(dataset=dict(pipeline=test_pipeline))
 val_dataloader = dict(dataset=dict(pipeline=test_pipeline))
 
-# training schedule for 1x
-train_cfg = dict(type='EpochBasedTrainLoop', max_epochs=12, val_interval=1)
+# training schedule for 6x
+max_epochs = 72
+train_cfg = dict(
+    type='EpochBasedTrainLoop', max_epochs=max_epochs, val_interval=5)
 val_cfg = dict(type='ValLoop')
 test_cfg = dict(type='TestLoop')
 
@@ -57,9 +59,9 @@
     dict(
         type='MultiStepLR',
         begin=0,
-        end=12,
+        end=max_epochs,
         by_epoch=True,
-        milestones=[8, 11],
+        milestones=[50],
         gamma=0.1)
 ]
 
@@ -68,3 +70,5 @@
     type='OptimWrapper',
     optimizer=dict(type='Adam', lr=2.5e-4),
     clip_grad=None)
+
+find_unused_parameters = True

demo/mono_det_demo.py

@@ -66,7 +66,7 @@ def main(args):
         wait_time=0,
         out_file=args.out_dir,
         pred_score_thr=args.score_thr,
-        vis_task='mono-det')
+        vis_task='mono_det')
 
 
 if __name__ == '__main__':

demo/multi_modality_demo.py

@@ -67,7 +67,7 @@ def main(args):
         wait_time=0,
         out_file=args.out_dir,
         pred_score_thr=args.score_thr,
-        vis_task='multi_modality-det')
+        vis_task='multi-modality_det')
 
 
 if __name__ == '__main__':

demo/pcd_demo.py

@@ -56,7 +56,7 @@ def main(args):
         wait_time=0,
         out_file=args.out_dir,
         pred_score_thr=args.score_thr,
-        vis_task='det')
+        vis_task='lidar_det')
 
 
 if __name__ == '__main__':

demo/pcd_seg_demo.py

@@ -51,7 +51,7 @@ def main(args):
         show=True,
         wait_time=0,
         out_file=args.out_dir,
-        vis_task='seg')
+        vis_task='lidar_seg')
 
 
 if __name__ == '__main__':

docs/zh_cn/user_guides/useful_tools.md

@@ -102,10 +102,10 @@ python tools/misc/browse_dataset.py configs/_base_/datasets/kitti-3d-3class.py -
 python tools/misc/browse_dataset.py configs/_base_/datasets/kitti-3d-3class.py --task det --aug --output-dir ${OUTPUT_DIR} --online
 ```
 
-如果您还想显示 2D 图像以及投影的 3D 边界框，则需要找到支持多模态数据加载的配置文件，然后将 `--task` 参数更改为 `multi_modality-det`。一个例子如下所示
+如果您还想显示 2D 图像以及投影的 3D 边界框，则需要找到支持多模态数据加载的配置文件，然后将 `--task` 参数更改为 `multi-modality_det`。一个例子如下所示
 
 ```shell
-python tools/misc/browse_dataset.py configs/mvxnet/mvxnet_fpn_dv_second_secfpn_8xb2-80e_kitti-3d-3class.py --task multi_modality-det --output-dir ${OUTPUT_DIR} --online
+python tools/misc/browse_dataset.py configs/mvxnet/mvxnet_fpn_dv_second_secfpn_8xb2-80e_kitti-3d-3class.py --task multi-modality_det --output-dir ${OUTPUT_DIR} --online
 ```
 
 ![](../../resources/browse_dataset_multi_modality.png)
@@ -121,7 +121,7 @@ python tools/misc/browse_dataset.py configs/_base_/datasets/scannet-seg.py --tas
 在单目 3D 检测任务中浏览 nuScenes 数据集
 
 ```shell
-python tools/misc/browse_dataset.py configs/_base_/datasets/nus-mono3d.py --task mono-det --output-dir ${OUTPUT_DIR} --online
+python tools/misc/browse_dataset.py configs/_base_/datasets/nus-mono3d.py --task mono_det --output-dir ${OUTPUT_DIR} --online
 ```
 
 ![](../../resources/browse_dataset_mono.png)

mmdet3d/datasets/__init__.py

@@ -1,6 +1,5 @@
 # Copyright (c) OpenMMLab. All rights reserved.
 from .builder import DATASETS, PIPELINES, build_dataset
-from .convert_utils import get_2d_boxes
 from .dataset_wrappers import CBGSDataset
 from .det3d_dataset import Det3DDataset
 from .kitti_dataset import KittiDataset
@@ -22,8 +21,8 @@
                          ObjectNameFilter, ObjectNoise, ObjectRangeFilter,
                          ObjectSample, PointSample, PointShuffle,
                          PointsRangeFilter, RandomDropPointsColor,
-                         RandomFlip3D, RandomJitterPoints, RandomShiftScale,
-                         VoxelBasedPointSampler)
+                         RandomFlip3D, RandomJitterPoints, RandomResize3D,
+                         RandomShiftScale, Resize3D, VoxelBasedPointSampler)
 from .utils import get_loading_pipeline
 from .waymo_dataset import WaymoDataset
 
@@ -40,5 +39,6 @@
     'LoadPointsFromMultiSweeps', 'WaymoDataset', 'BackgroundPointsFilter',
     'VoxelBasedPointSampler', 'get_loading_pipeline', 'RandomDropPointsColor',
     'RandomJitterPoints', 'ObjectNameFilter', 'AffineResize',
-    'RandomShiftScale', 'LoadPointsFromDict', 'PIPELINES', 'get_2d_boxes'
+    'RandomShiftScale', 'LoadPointsFromDict', 'PIPELINES',
+    'Resize3D', 'RandomResize3D',
 ]