resnet

This model tells you what's in an image. It's configured as a GPU example in cog.yaml.

It uses ResNet50 with ImageNet weights from torchvision. Torchvision fetches and caches the checkpoint the first time the model starts, so startup requires network access unless the checkpoint is already cached. Takes an image, returns the top-3 ImageNet classes.

Usage

First, make sure you've got the latest version of Cog installed.

Then run predictions on the model:

cog predict -i [email protected]