📍 Poster: Sat 10–12:30 PM, Hall 3 + 2B #516
💡LucidPPN clears up the confusion by separating color from shape and texture features, and by consistently aligning parts to the same object regions across classes. The result? More interpretable, more intuitive explanations — backed by user studies.
🔍 Prototypical parts models offer trustworthy but sometimes ambiguous explanations for their decisions…
I'm attending #ICLR #Singapore this week to present LucidPPN: Unambiguous Prototypical Parts Network for Interpretable Computer Vision. Let’s meet and chat explainable #AI 🙂
🧵👇
