Skip to content

Monocular 3D Detection FCOS3D Demo Failure due to size mismatch for neck.fpn_convs.3.conv.weight #719

@bml1g12

Description

@bml1g12

Describe the bug

I am using CUDA 11.1 on Ubunu 20.04 and docker:
docker run --gpus all --shm-size=8g -it mmdetection3d

Reproduction

Downloaded the finetuned model from here: https://download.openmmlab.com/mmdetection3d/v0.1.0_models/fcos3d/fcos3d_r101_caffe_fpn_gn-head_dcn_2x8_1x_nus-mono3d_finetune/fcos3d_r101_caffe_fpn_gn-head_dcn_2x8_1x_nus-mono3d_finetune_20210427_091419-35aaaad0.pth

And followed the demo here https://mmdetection3d.readthedocs.io/en/latest/demo.html#monocular-3d-detection:

root@d5a18d54d399:/mmdetection3d# python demo/mono_det_demo.py demo/data/nuscenes/n015-2018-07-24-11-22-45+0800__CAM_BACK__1532402927637525.jpg demo/data/nuscenes/n015-2018-07-24-11-22-45+0800__CAM_BACK__1532402927637525_mono3d.coco.json configs/fcos3d/fcos3d_r101_caffe_fpn_gn-head_dcn_2x8_1x_nus-mono3d_finetune.py checkpoints/fcos3d_r101_caffe_fpn_gn-head_dcn_2x8_1x_nus-mono3d_finetune_20210427_091419-35aaaad0.pth

/opt/conda/lib/python3.7/site-packages/mmcv/cnn/bricks/conv_module.py:107: UserWarning: ConvModule has norm and bias at the same time
  warnings.warn('ConvModule has norm and bias at the same time')
Use load_from_local loader
The model and loaded state dict do not match exactly

size mismatch for neck.fpn_convs.3.conv.weight: copying a param with shape torch.Size([256, 256, 3, 3]) from checkpoint, the shape in current model is torch.Size([256, 2048, 3, 3]).

Would be very grateful for assistance, as this is my first time using the repo and thought the demo would be a good place to start. Am I doing something wrong?

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions