Research map / Video / Video segmentation and tracking
ColorMNet: A Memory-based Deep Spatial-Temporal Feature Propagation Network for Video Colorization
Yixin Yang, Jiangxin Dong, Jinhui Tang, Jinshan Pan
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 |
|---|---|---|---|
| replace ResNet50 or DINOv2 features with PVGFE cross-attention fusion (ablation) | ColorMNet · DAVIS val | PSNR (dB): 35.01 -> 35.77 | Improved |
| fuse DINOv2 and ResNet50 via cross-attention instead of plain concatenation (ablation) | ColorMNet · DAVIS val | PSNR (dB): 35.26 -> 35.77 | Improved |
| use memory-based feature propagation instead of stacking all frames (ablation) | ColorMNet · DAVIS val | PSNR (dB): 33.94 -> 35.77 | Improved |
| use memory-based feature propagation instead of recurrent propagation (ablation) | ColorMNet · DAVIS val | PSNR (dB): 35.26 -> 35.77 | Improved |
| add local attention module (ablation) | ColorMNet · DAVIS val | PSNR (dB): 35.44 -> 35.77 | Improved |
| use MFP memory instead of stacking in terms of memory and speed (ablation) | ColorMNet · NVCC2023 val | Memory consumption (G): 24.1 -> 1.9 | Improved |
| MFP versus stacking running time (ablation) | ColorMNet · NVCC2023 val | Running time (s): 1.04 -> 0.07 | Improved |
| increase memory size parameters Ne, Ns and M (ablation) | MFP module · DAVIS val | PSNR (dB) | Mixed |
| change frame sampling interval gamma away from 5 (ablation) | MFP module · DAVIS val | PSNR (dB) | Hurt |
Later papers that compared with this method
1 later accepted paper reports colormnet as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- Many-for-Many: Unify the Training of Multiple Video and Image Generation and Manipulation Tasks (ICLR 2026 · better on 1 of 1 shared result)
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
- BiSTNet: Semantic Image Prior Guided Bidirectional Temporal Feature Fusion for Deep Exemplar-Based Video Colorization (TPAMI 2024 · similarity 0.88)
- Learned Variational Video Color Propagation (ECCV 2022 · similarity 0.88)
- Learning Representations for Automatic Colorization (ECCV 2016 · similarity 0.88)
- Deep Exemplar-Based Video Colorization (CVPR 2019 · similarity 0.88)
- ColorFormer: Image Colorization via Color Memory Assisted Hybrid-Attention Transformer (ECCV 2022 · similarity 0.87)
- LongAnimation: Long Animation Generation with Dynamic Global-Local Memory (ICCV 2025 · 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.
