Research map / Detection and segmentation / Open-set and few-shot detection
Deliberative Explanations: visualizing network insecurities
Pei Wang, Nuno Nvasconcelos
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Replace hesitancy score with hardness score (ablation) | difficulty score · ADE20K | ADE20K IoU precision (avg): 21.79(0.03) -> 23.80(0.19) | Improved |
| Replace entropy score with hardness score (ablation) | difficulty score · ADE20K | ADE20K IoU precision (avg): 20.73(0.09) -> 23.80(0.19) | Improved |
| Use second-order gradient-Hessian instead of first-order gradient (ablation) | attribution function · ADE20K | ADE20K IoU precision (avg): 23.80(0.19) -> 24.26(0.30) | No clear effect |
| Use second-order gradient-Hessian instead of first-order gradient (ablation) | attribution function · CUB200 | CUB200 precision-recall | Improved |
| Use integrated gradient instead of first-order gradient (ablation) | attribution function · ADE20K | ADE20K IoU precision (avg): 23.80(0.19) -> 23.99(0.24) | No clear effect |
| Add difficulty-score attribution map to gradient (vs gradient w/o ms) (ablation) | gradient attribution · ADE20K | ADE20K IoU precision (avg): 23.33(0.82) -> 23.80(0.19) | Improved |
| Use ResNet50 instead of VGG16 backbone (ablation) | deliberative explanations · ADE20K | ADE20K IoU precision (avg): 23.80(0.19) -> 22.60(0.22) | Hurt |
| Use ResNet50 instead of AlexNet backbone (ablation) | deliberative explanations · CUB200 | CUB200 precision-recall | Improved |
| Use ResNet50 instead of VGG16 backbone (ablation) | deliberative explanations · CUB200 | CUB200 precision-recall | Improved |
| Use VGG16 instead of AlexNet backbone (ablation) | deliberative explanations · ADE20K | ADE20K IoU precision (avg): 22.55(0.65) -> 23.80(0.19) | Improved |
| Show algorithm insecurities instead of random crops to turkers (ablation) | MTurk evaluation · MTurk | agreement with algorithm: 26.3% -> 51.9% | Improved |
| Show algorithm insecurities instead of random crops to turkers (ablation) | MTurk evaluation · MTurk | turker mutual agreement: 33.7% -> 59.4% | Improved |
Closest papers by meaning
- A Generalized Explanation Framework for Visualization of Deep Learning Model Predictions (TPAMI 2023 · similarity 0.89)
- Keep the Faith: Faithful Explanations in Convolutional Neural Networks for Case-Based Reasoning (AAAI 2024 · similarity 0.85)
- Towards Fine-Grained Interpretability: Counterfactual Explanations for Misclassification with Saliency Partition (CVPR 2025 · similarity 0.85)
- Comparing the Decision-Making Mechanisms by Transformers and CNNs via Explanation Methods (CVPR 2024 · similarity 0.85)
- Counterfactual Explanations for Misclassified Images: How Human and Machine Explanations Differ (Abstract Reprint) (AAAI 2024 · similarity 0.85)
- I-CEE: Tailoring Explanations of Image Classification Models to User Expertise (AAAI 2024 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
