pytorch-retinanet
pytorch-retinanet copied to clipboard
yhenon
Pytorch implementation of RetinaNet object detection.
**When the target frame is narrow and long, the regression loss is 0**！ `D:\YXC\Study\Anaconda\envs\pt-gpu\python.exe C:/Users/59593/Desktop/pytorch-retinanet-master/train.py CUDA available: True Num training images: 2 CUDA available: True CUDA available: True CUDA available:...
when i run the coco_validation.py, the error will arise~ Does anybody have this problem？ 
Has anyone tried to train coco2014 (or 2017 even) on Colab? I've got a tcmalloc: large alloc error after the first epoch. Is this related to GPU RAM ?
the nms doesn't according score ,why not according each class and score .
The class Anchors (in file retinanet/anchors.py) throws errors if anything other than the default "None" parameters are provided to it on initialization. In particular, I'm setting the pyramid levels, but...
Hi, I noticed one of issues before, your answer about freeze BN layer is about batchsize. My question is: According your code [here](https://github.com/yhenon/pytorch-retinanet/blob/1135e18b835481b18fd0d4e1613c87afc2bc7d46/model.py#L228). The function **freeze_bn** filter all BN layer...
## missing something file: train.py line: 126 ``` classification_loss, regression_loss = retinanet([data['img'].cuda().float(), data['annot']]) ``` I think `data['annot']` should be `data['annot'].cuda()`
Hi, thanks for your work. In your code, you use scheduler and set patience on training loss, is that correct? During traing, the model tend to overfit the training set,...
I have writen a test-file to valuate one image, But no matter what the image input is , the result is always None. So, I suspect my way of loading...
