lidless-vision

i have a private dataset of 300k semantically similar images. with the default model (5 resnet blocks, default settings. trained on gtx1060) i was able to achieve 1.75bpsp. i also...