arxivPaper only2026-06-08
Adversarial Attack and Disturbance Detection by Hadamard-Coded Output Representations for Object Detection and Semantic Segmentation
Conventional one-hot encodings often yield poorly calibrated models, being overconfident under attack, and let
品質予測/異常検知深層学習正規化・最適化手法分類検出セグメンテーション
- 用途
- 分類
- 難易度
- Hard
- コスト
- Low
→