potterhsu/easy-fpn.pytorch

An easy implementation of FPN (https://arxiv.org/pdf/1612.03144.pdf) in PyTorch.

easy-fpn.pytorch's Issues

精度好像没有提升多少

The accuracy problem

Hello, I find that the accuracy of easy-fpn has not significantly exceeded that of your easy-faster-rcnn you achieved much more. I want to know why? Looking forward to your reply

RuntimeError:cannot unsqueezze empty tensor

in training process,I got error below:
Traceback (most recent call last):
File "/media/liye/volume0/Ubuntu/PythonProject/easy-fpn.pytorch/train.py", line 150, in
main()
File "/media/liye/volume0/Ubuntu/PythonProject/easy-fpn.pytorch/train.py", line 148, in main
_train(dataset_name, backbone_name, path_to_data_dir, path_to_checkpoints_dir, path_to_resuming_checkpoint)
File "/media/liye/volume0/Ubuntu/PythonProject/easy-fpn.pytorch/train.py", line 52, in train
for batch_index, (, image_batch, _, bboxes_batch, labels_batch) in enumerate(dataloader):
File "/home/liye/anaconda3/envs/easy-fpn/lib/python3.6/site-packages/torch/utils/data/dataloader.py", line 336, in next
return self._process_next_batch(batch)
File "/home/liye/anaconda3/envs/easy-fpn/lib/python3.6/site-packages/torch/utils/data/dataloader.py", line 357, in _process_next_batch
raise batch.exc_type(batch.exc_msg)
RuntimeError: Traceback (most recent call last):
File "/home/liye/anaconda3/envs/easy-fpn/lib/python3.6/site-packages/torch/utils/data/dataloader.py", line 106, in _worker_loop
samples = collate_fn([dataset[i] for i in batch_indices])
File "/home/liye/anaconda3/envs/easy-fpn/lib/python3.6/site-packages/torch/utils/data/dataloader.py", line 187, in default_collate
return [default_collate(samples) for samples in transposed]
File "/home/liye/anaconda3/envs/easy-fpn/lib/python3.6/site-packages/torch/utils/data/dataloader.py", line 187, in
return [default_collate(samples) for samples in transposed]
File "/home/liye/anaconda3/envs/easy-fpn/lib/python3.6/site-packages/torch/utils/data/dataloader.py", line 164, in default_collate
return torch.stack(batch, 0, out=out)
RuntimeError: cannot unsqueeze empty tensor

Process finished with exit code 1

Supports Multi-Batch and Multi-GPU training and PyTorch 1?

请问loss一直在震荡是什么原因