|
|
--- |
|
|
license: apache-2.0 |
|
|
base_model: |
|
|
- QuanSun/EVA-CLIP |
|
|
pipeline_tag: image-to-image |
|
|
tags: |
|
|
- ReID |
|
|
- CC-ReID |
|
|
- LTCC |
|
|
- PRCC |
|
|
- CCVID |
|
|
- MeVID |
|
|
metrics: |
|
|
- accuracy |
|
|
--- |
|
|
|
|
|
# Colors See Colors Ignore: Clothes Changing ReID with Color Disentanglement (ICCV-25 🥳) |
|
|
|
|
|
Clothes-Changing Re-Identification (CC-ReID) aims to recognize individuals across different locations and times, irrespective of clothing. |
|
|
Our appporach "Colors See Colors Ignore (CSCI)" uses colors (lightweight, annotation-free) derived from RGB input to learn clothes robust embeddings. |
|
|
CSCI efficiently captures color-related appearance bias ('Color See') while disentangling it from identity-relevant ReID features ('Color Ignore'), all within 1 tranformer, with a novel self-attention prevent information leak between color and identity cues within the feature space. |
|
|
|
|
|
|
|
|
|
|
|
- Project Page: https://ucf-crcv.github.io/ReID/CSCI/ |
|
|
- GitHub Code : https://github.com/ppriyank/ICCV-CSCI-Person-ReID |
|
|
- Arxiv : https://arxiv.org/abs/2507.07230 |
|
|
- Paper : |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
