change the neck from SECONDFPN to FPN in the official config 'hv_pointpillars_secfpn_6x8_160e_kitti-3d-3class.py'.

[Noillness]

I tried to change the neck from SECONDFPN to FPN in the official config 'configs/pointpillars/hv_pointpillars_secfpn_6x8_160e_kitti-3d-3class.py' referring to 'configs/base/models/hv_pointpillars_fpn_nus.py', but the dimension of anchors is out of range in the function anchor_target_3d_single in train_mixins.py. I tried to debug it, it seemed that for FPN, the anchors should be a list of tensor, but it's a tensor in fact, so it entered a false if branch.

This is my config:
`base = [
# 'configs\pointpillars\hv_pointpillars_secfpn_6x8_160e_kitti-3d-3class.py'
'configs/pointpillars/hv_pointpillars_secfpn_6x8_160e_kitti-3d-3class.py'
]

voxel_size = [0.16, 0.16, 4]
model = dict(
type='VoxelNet',
voxel_layer=dict(
max_num_points=32, # max_points_per_voxel
point_cloud_range=[0, -39.68, -3, 69.12, 39.68, 1],
voxel_size=voxel_size,
max_voxels=(16000, 40000) # (training, testing) max_voxels
),
voxel_encoder=dict(
type='PillarFeatureNet',
in_channels=4,
feat_channels=[64],
with_distance=False,
voxel_size=voxel_size,
point_cloud_range=[0, -39.68, -3, 69.12, 39.68, 1]),
middle_encoder=dict(
type='PointPillarsScatter', in_channels=64, output_shape=[496, 432]),
backbone=dict(
type='SECOND',
in_channels=64,
layer_nums=[3, 5, 5],
layer_strides=[2, 2, 2],
out_channels=[64, 128, 256]),
# neck=dict(
# type='SECONDFPN',
# in_channels=[64, 128, 256],
# upsample_strides=[1, 2, 4],
# out_channels=[128, 128, 128]),
neck=dict(
delete=True,
type='FPN',
norm_cfg=dict(type='naiveSyncBN2d', eps=1e-3, momentum=0.01),
act_cfg=dict(type='ReLU'),
in_channels=[64, 128, 256],
out_channels=384,
# out_channels=256,
start_level=0,
num_outs=3),
bbox_head=dict(
type='Anchor3DHead',
num_classes=3,
# in_channels=256,
# feat_channels=256,
in_channels=384,
feat_channels=384,
use_direction_classifier=True,
assign_per_class=True,
anchor_generator=dict(
type='AlignedAnchor3DRangeGenerator',
ranges=[
[0, -39.68, -0.6, 69.12, 39.68, -0.6],
[0, -39.68, -0.6, 69.12, 39.68, -0.6],
[0, -39.68, -1.78, 69.12, 39.68, -1.78],
],
#
scales=[1, 2, 4],
#
sizes=[[0.8, 0.6, 1.73], [1.76, 0.6, 1.73], [3.9, 1.6, 1.56]],
rotations=[0, 1.57],
# reshape_out=False
reshape_out=True
),
diff_rad_by_sin=True,
bbox_coder=dict(type='DeltaXYZWLHRBBoxCoder'),
loss_cls=dict(
type='FocalLoss',
use_sigmoid=True,
gamma=2.0,
alpha=0.25,
loss_weight=1.0),
loss_bbox=dict(type='SmoothL1Loss', beta=1.0 / 9.0, loss_weight=2.0),
loss_dir=dict(
type='CrossEntropyLoss', use_sigmoid=False, loss_weight=0.2)),
# model training and testing settings
train_cfg=dict(
assigner=[
dict( # for Pedestrian
type='MaxIoUAssigner',
iou_calculator=dict(type='BboxOverlapsNearest3D'),
pos_iou_thr=0.5,
neg_iou_thr=0.35,
min_pos_iou=0.35,
ignore_iof_thr=-1),
dict( # for Cyclist
type='MaxIoUAssigner',
iou_calculator=dict(type='BboxOverlapsNearest3D'),
pos_iou_thr=0.5,
neg_iou_thr=0.35,
min_pos_iou=0.35,
ignore_iof_thr=-1),
dict( # for Car
type='MaxIoUAssigner',
iou_calculator=dict(type='BboxOverlapsNearest3D'),
pos_iou_thr=0.6,
neg_iou_thr=0.45,
min_pos_iou=0.45,
ignore_iof_thr=-1),
],
allowed_border=0,
pos_weight=-1,
debug=False),
test_cfg=dict(
use_rotate_nms=True,
nms_across_levels=False,
nms_thr=0.01,
score_thr=0.1,
min_bbox_size=0,
nms_pre=100,
max_num=50))

data = dict(
samples_per_gpu=2,
workers_per_gpu=1,
persistent_workers=True,
# train=dict(dataset=dict(ann_file='data/kitti/kitti_infos_val.pkl'), )
# test=dict(
# split='testing',
# ann_file='data/kitti/kitti_infos_test.pkl',
# )
)

optimizer = dict(
type='AdamW', lr=0.0001, betas=(0.95, 0.99), weight_decay=0.01)
lr_config = None
momentum_config = None

runner = dict(max_epochs=5)
checkpoint_config = dict(interval=5)
evaluation = dict(interval=5)
log_config = dict(interval=5)
`

This is my Traceback:
Traceback (most recent call last):
File "./tools/train.py", line 263, in
main()
File "./tools/train.py", line 252, in main
train_model(
File "/home/dingxiang/mmlab/mmdetection3d/mmdet3d/apis/train.py", line 344, in train_model
train_detector(
File "/home/dingxiang/mmlab/mmdetection3d/mmdet3d/apis/train.py", line 319, in train_detector
runner.run(data_loaders, cfg.workflow)
File "/home/dingxiang/anaconda3/envs/mmdetection3d/lib/python3.8/site-packages/mmcv/runner/epoch_based_runner.py", line 136, in run
epoch_runner(data_loaders[i], **kwargs)
File "/home/dingxiang/anaconda3/envs/mmdetection3d/lib/python3.8/site-packages/mmcv/runner/epoch_based_runner.py", line 53, in train
self.run_iter(data_batch, train_mode=True, **kwargs)
File "/home/dingxiang/anaconda3/envs/mmdetection3d/lib/python3.8/site-packages/mmcv/runner/epoch_based_runner.py", line 31, in run_iter
outputs = self.model.train_step(data_batch, self.optimizer,
File "/home/dingxiang/anaconda3/envs/mmdetection3d/lib/python3.8/site-packages/mmcv/parallel/data_parallel.py", line 77, in train_step
return self.module.train_step(*inputs[0], **kwargs[0])
File "/home/dingxiang/anaconda3/envs/mmdetection3d/lib/python3.8/site-packages/mmdet/models/detectors/base.py", line 248, in train_step
losses = self(**data)
File "/home/dingxiang/anaconda3/envs/mmdetection3d/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1102, in _call_impl
return forward_call(*input, **kwargs)
File "/home/dingxiang/anaconda3/envs/mmdetection3d/lib/python3.8/site-packages/mmcv/runner/fp16_utils.py", line 116, in new_func
return old_func(*args, **kwargs)
File "/home/dingxiang/mmlab/mmdetection3d/mmdet3d/models/detectors/base.py", line 60, in forward
return self.forward_train(**kwargs)
File "/home/dingxiang/mmlab/mmdetection3d/mmdet3d/models/detectors/voxelnet.py", line 94, in forward_train
losses = self.bbox_head.loss(
File "/home/dingxiang/anaconda3/envs/mmdetection3d/lib/python3.8/site-packages/mmcv/runner/fp16_utils.py", line 205, in new_func
return old_func(*args, **kwargs)
File "/home/dingxiang/mmlab/mmdetection3d/mmdet3d/models/dense_heads/anchor3d_head.py", line 342, in loss
cls_reg_targets = self.anchor_target_3d(
File "/home/dingxiang/mmlab/mmdetection3d/mmdet3d/models/dense_heads/train_mixins.py", line 71, in anchor_target_3d
neg_inds_list) = multi_apply(
File "/home/dingxiang/anaconda3/envs/mmdetection3d/lib/python3.8/site-packages/mmdet/core/utils/misc.py", line 30, in multi_apply
return tuple(map(list, zip(*map_results)))
File "/home/dingxiang/mmlab/mmdetection3d/mmdet3d/models/dense_heads/train_mixins.py", line 128, in anchor_target_3d_single
feat_size = anchors.size(0) * anchors.size(1) * anchors.size(2)
IndexError: Dimension out of range (expected to be in range of [-2, 1], but got 2)

[JingweiZhang12]

For pointpillars, we provide an example with FPN neck: https://github.com/open-mmlab/mmdetection3d/blob/master/configs/_base_/models/hv_pointpillars_fpn_nus.py

[Noillness]

I had referred to the config, my config is modified according to this config.I guessed the bug may be due to the diffrence between datasets. I used KITTI, but the configs with FPN all use nuScense.

[cat722]

Hello,I have the same problem. Have you solved it？