pytorch-retinanet
pytorch-retinanet copied to clipboard
yhenon
Pytorch implementation of RetinaNet object detection.
Can i train yellow format dataset in your model? YOLO format has txt file for each image.
Hi, I tried to train using CSV and there's a problem in csv_eval like this: IndexError: index -1 is out of bounds for axis 0 with size 0 in line...
Hi, I noticed that you used dataset instead of dataloader for evaluation. Why？
I wonder about the operation In the following lines: https://github.com/yhenon/pytorch-retinanet/blob/ef2c16f4ad66b652b836be36465ee63c978f3194/retinanet/dataloader.py#L363-L364 I assume that it suppose to round up the image dimensions to the closest number divisible by `32`. But, when...
What is the input size of the network, I found that there is a Resize class, which is (608, 1024), is the input size also (608, 1024)? But from the...
I'm re-training whole network from scratch with coco2017 training dataset and validation dataset. I ran 10 epochs over about 4 days but I'm having really low mAP. Like 0.102. I'm...
I tried to train the model using Resnet-50 as backbone. Here are the results I get after 18 epochs. Average Precision (AP) @[ IoU=0.50:0.95 | area= all | maxDets=100 ]...
Hi yhenon, thanks for your open source codes of retiannet. In my experiments, the result is much better than https://github.com/fizyr/keras-retinanet. I sincerely hope that you can implement the other loss...
The code uses BGR images (opencv reads images as BGR by default). Since training uses RGB images, inference should use RGB images as well.
