RuntimeError: Error(s) in loading state_dict for PointRCNN

[WakyWakyWayne]

RuntimeError: Error(s) in loading state_dict for PointRCNN:
size mismatch for rpn.backbone_net.SA_modules.0.mlps.0.layer0.conv.weight: copying a param with shape torch.Size([16, 3, 1, 1]) from checkpoint, the shape in current model is torch.Size([16, 4, 1,1])
size mismatch for rpn.backbone_net.SA_modules.0.mlps.1.layer0.conv.weight: copying a param with shape torch.Size([32, 3, 1, 1]) from checkpoint, the shape in current model is torch.Size([32, 4, 1, 1]).
size mismatch for rpn.backbone_net.FP_modules.0.mlp.layer0.conv.weight: copying a param with shape torch.Size([128, 256, 1, 1]) from checkpoint, the shape in current model is torch.Size([128, 257, 1, 1]).

[dmatos2012]

HI Waky,Wayky,

Did you download the PointRCNN.pth linked in the README?

[WakyWakyWayne]

Yes. I have done PointRCNN on Kitti, but I met this problem when I use your ros.

[gehadza]

set use_intentisity=False

[zhibingduan]

@gehadza @WakyWakyWayne Hello, I'm having the same problem and asked gpt about it saying it's due to a mismatch between the checkpoints and the model. You talk about changing use_intensity=false but I don't know where to change the value. By the way, if I retrain the model and run ros with the trained model, will it solve the problem? Any advice would be greatly appreciated!