pytorch-retinanet
pytorch-retinanet copied to clipboard
yhenon
Pytorch implementation of RetinaNet object detection.
Hi, thank you for your great work. I have a question about the format of the setting anchors. why in "shift_x = (np.arange(0, shape[1]) + 0.5) * stride" use shape[1]...
The original code doesn't consider the gpu situation for loss, and will lead to the problem: ```RuntimeError: Expected object of backend CUDA but got backend CPU for ...``` original code:...
Can someone clarify something for me? When training a model, are the weights initialized randomly or are the weights pre-trained on a well known dataset such as ImageNet? If the...
@yhenon Hi, thanks for sharing such nice code! I am confused about the annotation in https://github.com/yhenon/pytorch-retinanet/blob/1135e18b835481b18fd0d4e1613c87afc2bc7d46/model.py#L108. I think it should be "B x C x H x W" instead of...
https://github.com/yhenon/pytorch-retinanet/blob/a76e56eb537be476abc914587ae92ab542e2b3e4/train.py#L110 Can you provide a finetune code ??
I read the code but i don't understand how to calculate the loss for objects that have IOU
I am sorry to disturb you that I have a problem with the configuration environment. I have tried python2.7 pytorch0.4.1 as other issue says, but it didn't work. So what...
I am training the code on a custom CSV dataset, it works fine but where and how should I change to calculate the validation loss?
VisDrone is apparently a small dataset. As, I could perceive from these issues already opened is that retinanet does not perform well in the case of small datasets (Is there...
I changed the ratios from [0.5,1,2] to [0.25,0.5,1] in anchors.py at self.ratios (__init__()) and ratios (generate_anchors) . I get the followinf error: _### /pytorch/aten/src/ATen/native/cuda/IndexKernel.cu:60: operator(): block: [0,0,0], thread: [0,0,0] Assertion...
