AI Paper Atlas

Research map / Architectures and efficiency / CNN design

Deep Residual Learning in the JPEG Transform Domain

Max Ehrlich, Larry S. Davis

ICCV 2019 · Architectures and efficiency › CNN design · approach: Filter design and trained-filter analysis · 161 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
compute ReLu with Approximated Spatial Masking (ASM) instead of directly on the approximated block (APX) (ablation)JPEG-domain ReLu · random 8x8 blocks, 1-15 spatial frequenciesRMSEImproved
use ASM instead of APX ReLu in converted pretrained models (ablation)JPEG-domain ResNet via model conversion · MNIST, CIFAR-10, CIFAR-100test accuracyImproved
train in the JPEG domain rather than convert a spatial model, so fewer spatial frequencies sufficeJPEG-domain ResNet · MNIST, CIFAR-10, CIFAR-100test accuracyImproved
operate on JPEG coefficients instead of decompressed images for inferenceResNet · MNIST, CIFAR-10, CIFAR-100 (NVIDIA Pascal GPU, batch 40)inference throughputImproved
operate on JPEG coefficients instead of decompressed images for trainingResNet · MNIST, CIFAR-10, CIFAR-100 (NVIDIA Pascal GPU, batch 40)training throughputMixed

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy