pytorch-retinanet
pytorch-retinanet copied to clipboard
yhenon
Pytorch implementation of RetinaNet object detection.
I did not to use Parallel GPU by this code(torch.nn.DataParallel(retinanet).cuda()). So how to improve this code by using one gpu only? I appreciated you. Thank you.
How can we proceed further with the current implementation to use it for a localization task only. I have a custom dataset with no classification requirements as such. I just...
Hi, I'm getting Nan's after a few batches of training. First the bb predictions go to zero and later they flip to Nans, as do the losses. Does anybody have...
it seems to spend too much time in collater process when loading train data. so the GPU is waiting this . do you have good idea to solve it? thanks
Hello, I'm using your code and i managed to understand almost everything expect this line: `regression_loss = torch.where( torch.le(regression_diff, 1.0 / 9.0), 0.5 * 9.0 * torch.pow(regression_diff, 2), regression_diff -...
Traceback (most recent call last): File "train.py", line 173, in main() File "train.py", line 155, in main coco_eval.evaluate_coco(dataset_val, retinanet) File "/home/liuhao/seedland/retina/pytorch-retinanet/retinanet/coco_eval.py", line 73, in evaluate_coco coco_eval = COCOeval(coco_true, coco_pred, 'bbox')...
hi yhenon Thanks for providing in the implementation. I have below things to understand from you... 1) in Losses .py file why are we subtracting from width & height anchor[:,0]...
Thanks for your effective repos. You said > Currently, this repo achieves 33.7% mAP at 600px resolution with a Resnet-50 backbone. Unfortunately, for resnet-50 my model only reach 28.4% after...
出现一个bug：The expanded size of the tensor (120087) must match the existing size (20) at non-singleton dimension 1，请问输入的loc_preds和loc_targets中，anchors数量要一样吗？
