VocAlign Boosts Open‑Vocabulary Segmentation with Source‑Free Adaptation

VocAlign, a source‑free adaptation framework for vision‑language models, improves open‑vocabulary segmentation with a 6.11 mIoU gain on CityScapes using LoRA fine‑tuning. Read more: getnews.me/vocalign-boosts-open-voc... #vocalign #sourcefree