search

jimmyyhwu / pose-interpreter-networks

Real-time robotic object pose estimation with deep learning
MIT License
122 stars 27 forks source link

Problem in training the segmentation network using the provided dataset #33

Closed gfdeng closed 1 year ago

gfdeng commented 1 year ago

Hello, sorry to bother after so long, but using the dataset you provided, there was a problem training the segmentation network. The downloaded dataset is complete, the image in the OilChangeDataset/image has 9829 images. It runs perfectly when running visualize.ipynb and eval.ipynb. But there is a problem in the running train.py, I wonder if you know what the problem is?

Traceback (most recent call last): File "train_copy.py", line 238, in main(cfg) File "train_copy.py", line 202, in main train_batch_time, train_data_time, train_loss = train(train_loader, model, criterion, optimizer, epoch) File "train_copy.py", line 45, in trainfor i, (input, target) in enumerate(train_loader):

torch\utils\data\dataloader.py", line 530, in next data = self._next_data() torch\utils\data\dataloader.py", line 1224, in _next_data return self._process_data(data) torch\utils\data\dataloader.py", line 1250, in _process_data data.reraise() torch_utils.py", line 457, in reraise raise exception TypeError: Caught TypeError in DataLoader worker process 0.

jimmyyhwu commented 1 year ago

If you haven't already, I would try running the code with the package versions from August 2018 (e.g., pytorch 0.4, torchvision 0.2).

gfdeng commented 1 year ago

Thank you, this help me a lot.