Repo detail
TusharPyati04/Zero-Shot-Scene-Affordance-Detection-Using-Semantic-Embeddings-for-Open-World-Perception
This work proposes a zero-shot vision–language-based approach for drivable affordance detection. Using semantic embeddings and prompt-driven segmentation, it localizes navigable road regions without training and remains robust under diverse and adverse conditions.
Extracted labels
Project type: Library
Idea patterns: Dev tool
Scope: MVP
Audience: Public users
AI tools: Other
Confidence 0.90
Why these labels
- Focuses on affordance-centric scene understanding.
- Enables scalable and open-world perception.
- Robust under diverse and adverse conditions.
Commit activity (sampled)
Commits sampled
7
Active days
1
Build span
0 days
Median gap
— days
First commit
1/30/2026
Latest commit
1/30/2026
README keyword snippets
ction** in road scenes. The proposed approach leverages semantic embeddings and prompt-driven segmentation to localize navigable regions without task-specific training, enab