AI Paper Atlas

Research map / Video / Video segmentation and tracking

Overfitting the Data: Compact Neural Video Delivery via Content-Aware Feature Modulation

Jiaming Liu, Ming Lu, Kaixin Chen, Xiaoqi Li, Shizun Wang, Zhaoqing Wang, Enhua Wu, Yurong Chen, Chuang Zhang, Ming Wu

ICCV 2021 · Video › Video segmentation and tracking · approach: Learned video compression · 42 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
use content-aware (overfitted) training instead of external learningEDSR · Vid4 averagePSNR: 25.76 -> 28.32Improved
add CaFM with joint training, replacing separate per-chunk modelsEDSR-M · VSD4K game-45s x4PSNR: 34.32 -> 34.61Improved
add CaFM with joint training, replacing separate per-chunk modelsEDSR-M · VSD4K inter-45s x4PSNR: 32.67 -> 32.35Hurt
add CaFM with joint training, replacing separate per-chunk modelsEDSR-M · Vimeo-90K video 72549854 x3PSNR: 35.47 -> 35.40Hurt
add joint training instead of finetuning only CaFM on frozen M0 (ablation)EDSR-M CaFM · VSD4K game-15s x4PSNR: 33.49 -> 34.47Improved
use larger depthwise kernel in CaFM (1x1 to 7x7) (ablation)EDSR-M CaFM · VSD4K game-15s x4PSNR: 34.47 -> 34.54No clear effect
use 7x7 CaFM kernel instead of 1x1 (ablation)EDSR-M CaFM · EDSR parameter overheadparameter percentage: 0.60% -> 7.40%Hurt
apply CaFM joint training to VDSR backboneVDSR · VSD4K game-45s x4PSNR: 31.28 -> 33.16Improved
compare CaFM with H.265 at same storageEDSR-M CaFM · VSD4K game-45s x2PSNR: 39.77 -> 43.32Improved
compare CaFM with H.265 at same storageEDSR-M CaFM · VSD4K game-45s x4PSNR: 36.42 -> 34.61Hurt
use larger content-aware model (EDSR-L vs EDSR-M) (ablation)content-aware EDSR · Vid4 averagePSNR: 28.32 -> 29.88Improved

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