Research map / Architectures and efficiency / Quantization
Post-Training Quantization for Video Matting
Tianrui Zhu, Houyuan Chen, Ruihao Gong, Michele Magno, Haotong Qin, Kai Zhang
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 |
|---|---|---|---|
| add Global Affine Calibration (GAC) to BRECQ (ablation) | BRECQ · VM 512x288, W4A4 | Alpha MAD: 168.34 -> 50.75 | Improved |
| add GAC to QDrop (ablation) | QDrop · VM 512x288, W4A4 | Alpha MAD: 24.36 -> 22.01 | Improved |
| add Optical Flow Assistance (OFA) on top of BRECQ+GAC (ablation) | BRECQ+GAC · VM 512x288, W4A4 | Alpha MAD: 50.75 -> 46.16 | Improved |
| add OFA on top of BRECQ+GAC (ablation) | BRECQ+GAC · VM 512x288, W4A4 | DTSSD: 8.01 -> 3.15 | Improved |
| add OFA on top of QDrop+GAC (ablation) | QDrop+GAC · VM 512x288, W4A4 | Alpha MSE (x1e-3): 11.85 -> 13.80 | Hurt |
| add OFA on top of QDrop+GAC (ablation) | QDrop+GAC · VM 512x288, W4A8 | Alpha MSE (x1e-3): 4.43 -> 4.54 | No clear effect |
| use block-wise initial quantization (BIQ) instead of naive full-network quantization (ablation) | PTQ4VM · VM, W4A4 and W4A8 | Alpha error convergence | Improved |
| apply Cross-Layer Equalization and bias absorption before quantization | RVM PTQ · VM | accuracy | No clear effect |
| use PTQ4VM instead of QDrop at 4-bit weights | RVM · VM 512x288, W4A8 | Alpha MAD: 11.72 -> 10.77 | Improved |
| use PTQ4VM at W8A8 instead of FP32 | RVM · D646 512x512, W8A8 | Alpha MAD: 6.63 -> 7.14 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- QDrop — QDrop: Randomly Dropping Quantization for Extremely Low-bit Post-Training Quantization (ICLR 2022) · better on 6 of 6 shared results
- BRECQ — BRECQ: Pushing the Limit of Post-Training Quantization by Block Reconstruction (ICLR 2021) · better on 3 of 3 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- MatAnyone 2: Scaling Video Matting via a Learned Quality Evaluator (CVPR 2026 · similarity 0.90)
- VideoMaMa++: Temporally Consistent Video Matting via Preserve-and-Refine Tokens (NeurIPS 2026 · similarity 0.89)
- VideoMaMa: Mask-Guided Video Matting via Generative Prior (CVPR 2026 · similarity 0.89)
- MatAnyone: Stable Video Matting with Consistent Memory Propagation (CVPR 2025 · similarity 0.89)
- Unifying Automatic and Interactive Matting with Pretrained ViTs (CVPR 2024 · similarity 0.87)
- DSAQuant: Denoising-Stage-Aligned Quantization-Aware Training for Video Generation (NeurIPS 2026 · similarity 0.86)
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.
