change the backbone and the training is terminated in the process about bbavectors-oriented-object-detection HOT 6 OPEN

You can change the backbone, but make sure to output four feature maps to feed the decoder. The channels of the four feature maps are x[-1]=2048, x[-2]=1024, x[-3]=512, x[-4]=256. You may change the channels in lines14-16 of ctrbox_net.py, or delete some or add some of them to match your new backbone output.

from bbavectors-oriented-object-detection.

The backbone output can match and it can be trained, but it was interrupted as the error goes, so what may be wrong ?

from bbavectors-oriented-object-detection.

The exception looks like you don't have enough memory to fork child process? How about checking your /dev/shm(which I used before then I found I didn't have enough shared memory). Or decreasing the number of num_workers (set it to 0?)

from bbavectors-oriented-object-detection.

I have decreased the num_workers to 0, and it stills went wrong

from bbavectors-oriented-object-detection.

A different error?

from bbavectors-oriented-object-detection.

Yes, a different error:
File "BBAVectors-Oriented-Object-Detection-master/train.py", line 45, in save_model
}, path)
File "anaconda3/envs/maskrcnn/lib/python3.7/site-packages/torch/serialization.py", line 224, in save
return _with_file_like(f, "wb", lambda f: _save(obj, f, pickle_module, pickle_protocol))
File "anaconda3/envs/maskrcnn/lib/python3.7/site-packages/torch/serialization.py", line 149, in _with_file_like
return body(f)
File "anaconda3/envs/maskrcnn/lib/python3.7/site-packages/torch/serialization.py", line 224, in
return _with_file_like(f, "wb", lambda f: _save(obj, f, pickle_module, pickle_protocol))
File "anaconda3/envs/maskrcnn/lib/python3.7/site-packages/torch/serialization.py", line 303, in _save
serialized_storages[key]._write_file(f, _should_read_directly(f))
RuntimeError: std::bad_alloc

from bbavectors-oriented-object-detection.