Research map / Vision-language and multimodal / Visual question answering
U-CAM: Visual Explanation Using Uncertainty Based Class Activation Maps
Badri N. Patro, Mayank Lunayach, Shivansh Patel, Vinay P. Namboodiri
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 |
|---|---|---|---|
| add predictive uncertainty loss (PUL) to MCB (ablation) | MCB · VQA-v1 test-dev (ablation) | accuracy (All): 63.8 -> 64.9 | Improved |
| add aleatoric uncertainty loss (AUL) to MCB (ablation) | MCB · VQA-v1 test-dev (ablation) | accuracy (All): 63.8 -> 64.7 | Improved |
| add variance equalizer (VE) loss to MCB (ablation) | MCB · VQA-v1 test-dev (ablation) | accuracy (All): 63.8 -> 64.1 | Improved |
| add uncertainty distorted loss (UDL) to MCB (ablation) | MCB · VQA-v1 test-dev (ablation) | accuracy (All): 63.8 -> 64.4 | Improved |
| combine PUL and UDL instead of PUL alone (ablation) | MCB · VQA-v1 test-dev (ablation) | accuracy (All): 64.9 -> 65.9 | Improved |
| add UDL+VE on top of AUL+UDL (full A-GCA) (ablation) | MCB · VQA-v1 test-dev (ablation) | accuracy (All): 65.6 -> 66.3 | Improved |
| add VE and UDL on top of PUL (full P-GCA) (ablation) | MCB · VQA-v1 test-dev (ablation) | accuracy (All): 65.9 -> 66.5 | Improved |
| add PUL loss to MCB (ablation) | MCB · VQA-HAT (human attention) | rank correlation: 0.2790 -> 0.3012 | Improved |
| full P-GCA instead of baseline (ablation) | MCB · VQA-HAT (human attention) | rank correlation: 0.2790 -> 0.3341 | Improved |
| full P-GCA instead of baseline (ablation) | MCB · VQA-HAT (human attention) | EMD: 0.3931 -> 0.3721 | Improved |
| add VE loss to MCB (ablation) | MCB · VQA-HAT (human attention) | EMD: 0.3931 -> 0.3931 | No clear effect |
| add P-GCA to SAN | SAN · VQA-v2 test-dev | accuracy (All): SAN-2 56.9 -> 59.2 | Mixed |
Closest papers by meaning
- Explanation vs Attention: A Two-Player Game to Obtain Attention for VQA (AAAI 2020 · similarity 0.90)
- Self-Critical Reasoning for Robust Visual Question Answering (NeurIPS 2019 · similarity 0.88)
- Re-Attention for Visual Question Answering (AAAI 2020 · similarity 0.88)
- Grad-CAM: Visual Explanations From Deep Networks via Gradient-Based Localization (ICCV 2017 · similarity 0.88)
- An Analysis of Visual Question Answering Algorithms (ICCV 2017 · similarity 0.88)
- FOCUS: Internal MLLM Representations for Efficient Fine-Grained Visual Question Answering (NeurIPS 2025 · similarity 0.87)
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.
