GoGoDuck912/Self-Correction-Human-Parsing
A state-of-the-art human parsing model that segments human figures into 20 body parts and clothing categories.
Velocity · 7d
+0.5
★ / day
Trend
→steady
star history
The repository provides pretrained deep learning models and inference code for semantic segmentation of human bodies. It includes training pipelines and out-of-box extraction tools for downstream applications. The models achieve 59.36% mIoU on LIP and 82.29% mIoU on ATR benchmarks, with extensions for multi-person and video scenarios.