AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

"L3: Accelerator-Friendly Lossless Image Format for High-Resolution, High-Throughput DNN Training"

Jonghyun Bae, Woohyeon Baek, Tae Jun Ham, Jae W. Lee

ECCV 2022 · Architectures and efficiency › RNNs and normalization · approach: Differentiable optimization layers · 1 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
add patch-level parallelism to the L3 decoder (ablation)L3 decoder, Cityscapes · HD/FHD/UHDdecode time: reduces the decoding time by 76.8%, 83.9%, and 86.5% for HD, FHD, and UHDImproved
add pixel-wise parallel base-delta decoding (ablation)L3 decoder, Cityscapes · HD/FHD/UHDdecode time: reduces the decoding time by an average of 10.8%Improved
use the custom Paeth filter (top-left, top, top-right) with row-wise parallel decode (ablation)L3 decoder, Cityscapes · HD/FHD/UHDdecode time: reduces the decoding time by an average of 46.0%Improved
combine pixel-wise base-delta decoding and custom Paeth filter (ablation)L3 decoder, Cityscapes · HD/FHD/UHDdecode time: 49.5%, 56.7%, and 59.1% from the Baseline for HD, FHD, and UHDImproved
replace PNG with L3 for lossless data preparationDNN training pipeline · Cityscapes HD/FHD/UHDdata preparation throughput: 5.67x, 9.29x, and 15.71x for HD, FHD, and UHDImproved
replace lossless WebP with L3DNN training pipeline · Cityscapes HD/FHD/UHDdata preparation throughput: 2.41x, 3.25x, and 4.51xImproved
replace PNG with L3end-to-end training of seven models · Cityscapes and KITTItraining throughput: 1.71x geomean speedupImproved
replace lossy JPEG with L3 at equal accuracyend-to-end training of seven models · Cityscapes and KITTItraining throughput: 1.25x geomean, up to 1.77xImproved
replace lossy WebP with L3 at equal accuracyend-to-end training of seven models · Cityscapes and KITTItraining throughput: 1.93x geomean, up to 2.87xImproved
use default-quality JPEG (Q75) instead of lossless PNGDNN training on PNG vs JPEG · DDRNet23-slim, CityscapesmIoU: 0.729 -> 0.693Hurt
use default-quality JPEG (Q75) instead of lossless PNGDNN training on PNG vs JPEG · MaskFormer, CityscapesmIoU: 0.685 -> 0.669Hurt
use JPEG (Q75) instead of PNGDeepLabv3+ · CityscapesmIoU: 0.801 -> 0.801No clear effect

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