I wanted to try and play with the object detection torchvision pretrained models for a personal application in which I need to detect objects in pictures. In particular, for efficiency reasons, I need to perform batch inference. However, I noticed that the torchvision.models do not support batch inference, indeed, they simply take a list of tensors as input and they iterate over the list.
What am I missing?
In my case, I’m using retinanet_resnet50_fpn, but I believe that the aforementioned problem applies also for the other detection models.
Thank you very much.