update task name and config for fcos3d++ on nus (#1927)

configs/_base_/datasets/waymoD5-mono3d-3class.py

@@ -81,7 +81,7 @@ train_dataloader = dict(
         # we use box_type_3d='LiDAR' in kitti and nuscenes dataset
         # and box_type_3d='Depth' in sunrgbd and scannet dataset.
         box_type_3d='Camera',
-        task='mono3d',
+        task='mono_det',
         # load one frame every three frames
         load_interval=5))
 
@@ -109,7 +109,7 @@ val_dataloader = dict(
         # we use box_type_3d='LiDAR' in kitti and nuscenes dataset
         # and box_type_3d='Depth' in sunrgbd and scannet dataset.
         box_type_3d='Camera',
-        task='mono3d',
+        task='mono_det',
     ))
 
 test_dataloader = dict(
@@ -136,7 +136,7 @@ test_dataloader = dict(
         # we use box_type_3d='LiDAR' in kitti and nuscenes dataset
         # and box_type_3d='Depth' in sunrgbd and scannet dataset.
         box_type_3d='Camera',
-        task='mono3d',
+        task='mono_det',
     ))
 
 val_evaluator = dict(
@@ -145,5 +145,5 @@ val_evaluator = dict(
     waymo_bin_file='./data/waymo/waymo_format/cam_gt.bin',
     data_root='./data/waymo/waymo_format',
     metric='LET_mAP',
-    task='mono3d')
+    task='mono_det')
 test_evaluator = val_evaluator

configs/pgd/pgd_r101_fpn-head_dcn_16xb3_waymoD5-mono3d.py

+_base_ = [
+    '../_base_/datasets/waymoD5-mono3d-3class.py', '../_base_/models/pgd.py',
+    '../_base_/schedules/mmdet-schedule-1x.py', '../_base_/default_runtime.py'
+]
+# model settings
+model = dict(
+    backbone=dict(
+        type='mmdet.ResNet',
+        depth=101,
+        num_stages=4,
+        out_indices=(0, 1, 2, 3),
+        frozen_stages=1,
+        norm_cfg=dict(type='BN', requires_grad=True),
+        norm_eval=True,
+        style='pytorch',
+        init_cfg=dict(type='Pretrained', checkpoint='torchvision://resnet101'),
+        dcn=dict(type='DCNv2', deform_groups=1, fallback_on_stride=False),
+        stage_with_dcn=(False, False, True, True)),
+    neck=dict(num_outs=3),
+    bbox_head=dict(
+        num_classes=3,
+        bbox_code_size=7,
+        pred_attrs=False,
+        pred_velo=False,
+        pred_bbox2d=True,
+        use_onlyreg_proj=True,
+        strides=(8, 16, 32),
+        regress_ranges=((-1, 128), (128, 256), (256, 1e8)),
+        group_reg_dims=(2, 1, 3, 1, 16,
+                        4),  # offset, depth, size, rot, kpts, bbox2d
+        reg_branch=(
+            (256, ),  # offset
+            (256, ),  # depth
+            (256, ),  # size
+            (256, ),  # rot
+            (256, ),  # kpts
+            (256, )  # bbox2d
+        ),
+        centerness_branch=(256, ),
+        loss_cls=dict(
+            type='mmdet.FocalLoss',
+            use_sigmoid=True,
+            gamma=2.0,
+            alpha=0.25,
+            loss_weight=1.0),
+        loss_bbox=dict(
+            type='mmdet.SmoothL1Loss', beta=1.0 / 9.0, loss_weight=1.0),
+        loss_dir=dict(
+            type='mmdet.CrossEntropyLoss', use_sigmoid=False, loss_weight=1.0),
+        loss_centerness=dict(
+            type='mmdet.CrossEntropyLoss', use_sigmoid=True, loss_weight=1.0),
+        use_depth_classifier=True,
+        depth_branch=(256, ),
+        depth_range=(0, 50),
+        depth_unit=10,
+        division='uniform',
+        depth_bins=6,
+        pred_keypoints=True,
+        weight_dim=1,
+        loss_depth=dict(
+            type='UncertainSmoothL1Loss', alpha=1.0, beta=3.0,
+            loss_weight=1.0),
+        loss_bbox2d=dict(
+            type='mmdet.SmoothL1Loss', beta=1.0 / 9.0, loss_weight=0.0),
+        loss_consistency=dict(type='mmdet.GIoULoss', loss_weight=0.0),
+        bbox_coder=dict(
+            type='PGDBBoxCoder',
+            base_depths=((41.01, 18.44), ),
+            base_dims=(
+                (4.73, 1.77, 2.08),
+                (0.91, 1.74, 0.84),
+                (1.81, 1.77, 0.84),
+            ),
+            code_size=7)),
+    # set weight 1.0 for base 7 dims (offset, depth, size, rot)
+    # 0.2 for 16-dim keypoint offsets and 1.0 for 4-dim 2D distance targets
+    train_cfg=dict(code_weight=[
+        1.0, 1.0, 0.2, 1.0, 1.0, 1.0, 1.0, 0.2, 0.2, 0.2, 0.2, 0.2, 0.2, 0.2,
+        0.2, 0.2, 0.2, 0.2, 0.2, 0.2, 0.2, 0.2, 0.2, 1.0, 1.0, 1.0, 1.0
+    ]),
+    test_cfg=dict(nms_pre=100, nms_thr=0.05, score_thr=0.001, max_per_img=20))
+
+# optimizer
+optim_wrapper = dict(
+    optimizer=dict(
+        type='SGD',
+        lr=0.008,
+    ),
+    paramwise_cfg=dict(bias_lr_mult=2., bias_decay_mult=0.),
+    clip_grad=dict(max_norm=35, norm_type=2))
+
+param_scheduler = [
+    dict(
+        type='LinearLR',
+        start_factor=1.0 / 3,
+        by_epoch=False,
+        begin=0,
+        end=500),
+    dict(
+        type='MultiStepLR',
+        begin=0,
+        end=24,
+        by_epoch=True,
+        milestones=[16, 22],
+        gamma=0.1)
+]
+total_epochs = 24
+runner = dict(max_epochs=total_epochs)
+train_cfg = dict(type='EpochBasedTrainLoop', max_epochs=24, val_interval=24)
+val_cfg = dict(type='ValLoop')
+test_cfg = dict(type='TestLoop')

mmdet3d/datasets/waymo_dataset.py

@@ -93,7 +93,6 @@ class WaymoDataset(KittiDataset):
         self.cat_ids = range(len(self.METAINFO['CLASSES']))
         self.cat2label = {cat_id: i for i, cat_id in enumerate(self.cat_ids)}
         self.max_sweeps = max_sweeps
-        self.task = task
         # we do not provide file_client_args to custom_3d init
         # because we want disk loading for info
         # while ceph loading for KITTI2Waymo
@@ -108,6 +107,7 @@ class WaymoDataset(KittiDataset):
             default_cam_key=default_cam_key,
             data_prefix=data_prefix,
             test_mode=test_mode,
+            task=task,
             **kwargs)
 
     def parse_ann_info(self, info: dict) -> dict:
@@ -221,7 +221,7 @@ class WaymoDataset(KittiDataset):
 
                 # TODO check if need to modify the sample id
                 # TODO check when will use it except for evaluation.
-                camera_info['sample_idx'] = info['sample_idx']
+                camera_info['sample_id'] = info['sample_id']
 
                 if not self.test_mode:
                     # used in training

mmdet3d/evaluation/metrics/waymo_metric.py

@@ -67,7 +67,7 @@ class WaymoMetric(KittiMetric):
                  prefix: Optional[str] = None,
                  pklfile_prefix: str = None,
                  submission_prefix: str = None,
-                 task='lidar',
+                 task='lidar_det',
                  default_cam_key: str = 'CAM_FRONT',
                  use_pred_sample_idx: bool = False,
                  collect_device: str = 'cpu',
@@ -106,7 +106,7 @@ class WaymoMetric(KittiMetric):
         self.data_infos = load(self.ann_file)['data_list']
         # different from kitti, waymo do not need to convert the ann file
         # handle the mono3d task
-        if self.task == 'mono3d':
+        if self.task == 'mono_det':
             new_data_infos = []
             for info in self.data_infos:
                 height = info['images'][self.default_cam_key]['height']
@@ -461,7 +461,7 @@ class WaymoMetric(KittiMetric):
                 # If you want to use another camera, please modify it.
                 image_shape = (info['images'][self.default_cam_key]['height'],
                                info['images'][self.default_cam_key]['width'])
-            if self.task == 'mono3d':
+            if self.task == 'mono_det':
                 box_dict_per_frame.append(box_dict)
                 if (idx + 1) % self.num_cams != 0:
                     continue
@@ -600,9 +600,9 @@ class WaymoMetric(KittiMetric):
                 sample_idx=sample_idx)
         # Here default used 'CAM2' to compute metric. If you want to
         # use another camera, please modify it.
-        if self.task in ['mv3d', 'lidar']:
+        if self.task in ['mv3d_det', 'lidar_det']:
             cam_key = self.default_cam_key
-        elif self.task == 'mono3d':
+        elif self.task == 'mono_det':
             cam_key = list(info['images'].keys())[0]
         else:
             raise NotImplementedError
@@ -635,12 +635,12 @@ class WaymoMetric(KittiMetric):
                           (box_2d_preds[:, 1] < image_shape[0]) &
                           (box_2d_preds[:, 2] > 0) & (box_2d_preds[:, 3] > 0))
         # check box_preds_lidar
-        if self.task in ['lidar', 'mono3d']:
+        if self.task in ['mv3d_det', 'lidar_det']:
             limit_range = box_preds.tensor.new_tensor(self.pcd_limit_range)
             valid_pcd_inds = ((box_preds_lidar.center > limit_range[:3]) &
                               (box_preds_lidar.center < limit_range[3:]))
             valid_inds = valid_pcd_inds.all(-1)
-        elif self.task == 'mono3d':
+        elif self.task == 'mono_det':
             valid_inds = valid_cam_inds
 
         if valid_inds.sum() > 0: