AI Paper Atlas

Research map / Probabilistic and scientific modeling / Variational and generative sampling

Bit Allocation using Optimization

Tongda Xu, Han Gao, Chenjian Gao, Yuanyuan Wang, Dailan He, Jinyong Pi, Jixiang Luo, Ziyu Zhu, Mao Ye, Hongwei Qin, Yan Wang, Jingjing Liu and 1 more

ICML 2023 · Probabilistic and scientific modeling › Variational and generative sampling · approach: Amortized and online variational inference · 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
Use proposed pixel-level SAVI bit allocation instead of OEUDVC and DCVC · HEVC Class DBD-PSNR: DVC: 0.83 -> 1.67 dB; DCVC: 0.74 -> 1.55 dBImproved
Replace Proposed-Approx with the scalable approximation (ablation)DVC · HEVC Class DBD-BR: -35.86% -> -29.81%Hurt
Use scalable approximation instead of Proposed-Approx (ablation)DVC · HEVC Class Dencoding time: 528.3s -> 74.6sImproved
Optimize each frame independently rather than using GoP-level optimization (ablation)Original SAVI on DVC · HEVC Class DBD-BR: Original SAVI: -14.76%; Original SAVI (per-frame): -20.12%Improved
Reduce gradient-ascent steps from 2000 to 1000 (ablation)Proposed-Scalable on DVC · HEVC Class DBD-BR: -29.81 -> -25.37Hurt
Reduce gradient-ascent steps from 2000 to 500 (ablation)Proposed-Approx on DVC · HEVC Class DBD-BR: -35.86 -> -25.60Hurt
Use Original SAVI instead of FAVI initialization alone (ablation)2-level VAE · MNISTNLL lowerbound: ≤ 98.3113 -> ≤ 91.5530Improved
Use proposed approximate SAVI instead of Original SAVI (ablation)2-level VAE · MNISTNLL lowerbound: ≤ 91.5530 -> ≤ 91.5486Improved
Use accurate SAVI rather than approximate SAVI (ablation)2-level VAE · MNISTNLL lowerbound: ≤ 91.5486 -> ≤ 91.5482Improved
Restrict gradient likelihood evaluation to C future frames (ablation)Proposed-Scalable · NVCRAM scaling with GoP sizeImproved

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