This repository contains the code, experiments, and results accompanying the paper "Learned Compression for Compressed Learning"
RD: Evaluation of rate, distortion, perception, computation and dimension of audio and image codecs
train: Experiments and results for training compressed-domain image models (classification, colorization)
train_audio: Experiments and results for training audio models (music source separation)
FIC: Experiments and results for full-input compression via pre-trained models (document understanding)