Research map / Video / Video segmentation and tracking
VCT: A Video Compression Transformer
Fabian Mentzer, George Toderici, David Minnen, Sergi Caelles, Sung Jin Hwang, Mario Lucic, Eirikur Agustsson
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 changed | Where | Measured | Effect |
|---|---|---|---|
| condition entropy model on 1 previous frame instead of none (ablation) | VCT · MCL-JCV | bpp: 0.218 -> 0.0907 | Improved |
| use 2 previous frames instead of 1 (ablation) | VCT · MCL-JCV | bpp: 0.0907 -> 0.0775 | Improved |
| add latent residual predictor (LRP) (ablation) | VCT · MCL-JCV | PSNR: 36.1 -> 36.8 | Improved |
| use more than 2 previous frames of context (ablation) | VCT · MCL-JCV | bpp | No clear effect |
| replace flow/warping with transformer temporal entropy model | SSF-style video codec · Synthetic Shift videos | R-D loss: approx 45% lower R-D loss | Improved |
| transformer instead of built-in motion priors, static video | VCT · Synthetic static video (x=0) | R-D loss | Hurt |
| increase training set size to 50k+ clips (ablation) | VCT vs SSF · MCL-JCV | evaluation loss | Improved |
Closest papers by meaning
- Insights From Generative Modeling for Neural Video Compression (TPAMI 2023 · similarity 0.90)
- Transformer-based Transform Coding (ICLR 2022 · similarity 0.89)
- Video Compression With Entropy-Constrained Neural Representations (CVPR 2023 · similarity 0.89)
- C3: High-Performance and Low-Complexity Neural Compression from a Single Image or Video (CVPR 2024 · similarity 0.88)
- Towards Practical Real-Time Neural Video Compression (CVPR 2025 · similarity 0.87)
- TTVidT: Decoupling the Temporal Axis for Efficient Motion-Centric Video Pretraining (NeurIPS 2026 · similarity 0.87)
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.
