candle-resnet

A candle implementation of inference using a pre-trained ResNet. This uses a classification head trained on the ImageNet dataset and returns the probabilities for the top-5 classes.

Running an example

$ cargo run --example resnet --release -- --image tiger.jpg

loaded image Tensor[dims 3, 224, 224; f32]
model built
tiger, Panthera tigris  : 90.21%
tiger cat               : 8.93%
lion, king of beasts, Panthera leo: 0.35%
leopard, Panthera pardus: 0.16%
jaguar, panther, Panthera onca, Felis onca: 0.09%