Research map / Video / Video segmentation and tracking
Learning Binary Residual Representations for Domain-Specific Video Streaming
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Add a binary residual autoencoder to H.264 streaming | H.264 domain-specific video streaming · Four datasets, 5Mbps | PSNR: 1.71dB better than H.264 | Helped |
| Use the residual autoencoder rather than an artifact-removal network | Domain-specific video streaming · Four datasets, 5Mbps | PSNR: 0.84dB better | Helped |
| Add the binary residual representation to H.264 | Domain-specific video streaming · Four datasets, 2Mbps | SSIM: 5.3% | Helped |
| Use a binary autoencoder to compress residuals instead of sending raw-video autoencoder codes (ablation) | Video compression · KITTI, H.264 at 5Mbps | PSNR: 0.52dB drop | Helped |
| Use hardtanh or tanh binarization rather than sigmoid, stochastic regularization, or Gumbel-Sigmoid (ablation) | Binary residual autoencoder · KITTI, H.264 at 1M, 2M, and 5M | PSNR: 28.95, 29.97, 30.67 vs 28.89, 29.88, 30.61; 28.26, 29.21, 29.78; 26.66, 27.82, 28.66 | Helped |
| Use hardtanh instead of sigmoid, stochastic regularization, or Gumbel-Sigmoid (ablation) | Binary residual autoencoder · KITTI, H.264 at 1M, 2M, and 5M | SSIM: 0.8575, 0.8773, 0.8901 vs 0.8538, 0.8722, 0.8861; 0.8297, 0.8470, 0.8617; 0.7683, 0.8128, 0.8464 | Helped |
| Increase depth and channel count from (C, L)=(8, 3) to (32, 4) (ablation) | Binary residual autoencoder · KITTI, approximately 0.24Mbps residual stream | PSNR: 29.28 -> 29.39 | Helped |
| Increase depth and channel count from (C, L)=(16, 3) to (64, 4) (ablation) | Binary residual autoencoder · KITTI, approximately 0.48Mbps residual stream | PSNR: 29.83 -> 29.81 | No clear effect |
| Increase depth and channel count from (C, L)=(8, 2) to (32, 3) (ablation) | Binary residual autoencoder · KITTI, approximately 0.96Mbps residual stream | PSNR: 30.35 -> 30.67 | Helped |
Closest papers by meaning
- DeNC: Unleash Neural Codecs in Video Streaming with Diffusion Enhancement (AAAI 2025 · similarity 0.84)
- Hybrid-Level Instruction Injection for Video Token Compression in Multi-modal Large Language Models (CVPR 2025 · similarity 0.83)
- 4DGCPro: Efficient Hierarchical 4D Gaussian Compression for Progressive Volumetric Video Streaming (NeurIPS 2025 · similarity 0.83)
- One Token per Highly Selective Frame: Towards Extreme Compression for Long Video Understanding (NeurIPS 2025 · similarity 0.82)
- Extending Video Masked Autoencoders to 128 frames (NeurIPS 2024 · similarity 0.82)
- Blind Bitstream-corrupted Video Recovery via Metadata-guided Diffusion Model (CVPR 2025 · similarity 0.82)
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.