Faster RCNN -- RuntimeError: inconsistent tensor size at /py/conda-bld/pytorch_1493676237139/work/torch/lib/TH/generic/THTensorCopy.c:51

vision
1

getting above error upon loading faster rcnn trained model:
model is trained with faster rcnn with input 704X800X3 grid. Three channels in input are not RGB but it holds some real number values
training was successful and model is saved as faster_rcnn_1000.h5 but when I try to load model for
testing network.load_net(model_file, detector) throws above error

I am using https://github.com/longcw/faster_rcnn_pytorch version of faster_rcnn and tweaked it for my input and use case. I am new to Pytorch will really appreciate if someone can help and give insight

2

the answer is not obvious. open an issue on the particular repository.

3

[SOLVED] Thanks for reply…I later realized my mistake that in original implementation faster rcnn fc7 layer output 21 output values… but since I was training my model with 4 categories (reason for size mismatch) so now with little modification in faster rcnn network it is executing correctly.