Research map / Vision-language and multimodal / Human and embodied modeling
Ambiguous Images With Human Judgments for Robust Visual Event Classification
Kate Sanders, Reno Kriz, Anqi Liu, Benjamin Van Durme
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 |
|---|---|---|---|
| train on SQUID-E ambiguous images instead of standard high-certainty images | ResNet50 · SQUID-E 4-way event classification | avg accuracy: .61 -> .70 | Improved |
| train on SQUID-E instead of standard data, evaluated on standard data | ResNet50 · Standard data | accuracy: .93 -> .71 | Hurt |
| add geometric augmentation to standard-data training | ResNet50 · SQUID-E 4-way event classification | avg accuracy: .61 -> .64 | Improved |
| add noise injection and masking to standard-data training | ResNet50 · SQUID-E 4-way event classification | avg accuracy: .61 -> .64 | Improved |
| add photometric augmentation to standard-data training | ResNet50 · SQUID-E 4-way event classification | avg accuracy: .61 -> .57 | Hurt |
| add AugMix to standard-data training | ResNet50 · SQUID-E 4-way event classification | avg accuracy: .61 -> .60 | No clear effect |
| add combined augmentations to standard-data training | ResNet50 · SQUID-E 4-way event classification | standard data accuracy: .93 -> .87 | Hurt |
| use focal loss instead of cross-entropy baseline | ResNet50 trained on SD · SQUID-E validation | MSE vs human judgments: .15 -> .11 | Improved |
| use focal loss instead of cross-entropy baseline | ResNet50 trained on SD · SQUID-E validation | ECE: .58 -> .42 | Improved |
| use label smoothing instead of cross-entropy baseline | ResNet50 trained on SD · SQUID-E validation | ECE: .58 -> .46 | Improved |
| use Monte-Carlo Dropout instead of baseline | ResNet50 trained on SD · SQUID-E validation | MSE vs human judgments: .15 -> .14 | No clear effect |
| use focal loss for model trained on SQUID-E | ResNet50 trained on SQUID-E · SQUID-E validation | MSE vs human judgments: .16 -> .12 | Improved |
Closest papers by meaning
- Evaluating Multiview Object Consistency in Humans and Image Models (NeurIPS 2024 · similarity 0.85)
- Object Detection Benchmarks are Incomplete: The Role of Label Errors and Annotation Uncertainty (NeurIPS 2026 · similarity 0.84)
- Benchmarking Uncertainty Disentanglement: Specialized Uncertainties for Specialized Tasks (NeurIPS 2024 · similarity 0.83)
- Cocoon: Robust Multi-Modal Perception with Uncertainty-Aware Sensor Fusion (ICLR 2025 · similarity 0.83)
- LAION-C: An Out-of-Distribution Benchmark for Web-Scale Vision Models (ICML 2025 · similarity 0.83)
- Contour Integration Underlies Human-Like Vision (ICML 2025 · similarity 0.83)
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.
