rxqy

there might some difference between these models ie, maybe caffe pretrain model requires input normalization. while the origin one doesn't

Try running the code by adding `os.environ['CUDA_VISIBLE_DEVICES'] = '2' ` in train.py or `CUDA_VISIBLE_DEVICES = 2 python train.py xxxxxx` Hope this might help?

My guess is that there are some other places you might need to modify. i.e the RCNN head for final predictions (number of predictions in your error?) so it would...

Which version of pytorch are you using? If you are using pytorch 0.4, you need to git checkout to the 0.4 branch [https://github.com/CharlesShang/DCNv2#note](url)

Hi @dasing , I'm also reading into the same paper and find the code here. Here's my understanding of the equations above, correct me if I'm wrong? First, for simplicity...

Hi @dasing , for an arbitrary point, we should use equation 11 in the original paper. That is, for a tangent plane T ( tangent to the sphere in (theta_pie,...

Hi, I'll be using nvidia-ml-py 12.535.77 for now. Many thanks for the help.

Hi, a simple solution would be to add a .copy() to all of the output of rotate_img From https://github.com/gidariss/FeatureLearningRotNet/blob/817342ba8521ef709467c8bb7e3ec157969c1145/dataloader.py#L216-L226 To return np.flipud(...),copy() return np.fliplr(...).copy() return np.transpose(...).copy() For more details of...

Hi, I also encountered a similar problem here. Any suggestions?

Hi @NivekT , the above scipt works fine. And I did a little digging. This works and gives the correct results with ``` python -m torch.distributed.launch --nproc_per_node=2 --master_port 29100 fullsync.py...