Research map / Detection and segmentation / Open-set and few-shot detection
Expecting the Unexpected: Training Detectors for Unusual Pedestrians With Adversarial Imposters
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 adversarial imposters during adaptation, then fine-tune on real target images (ablation) | RPN+ pretrained on synthetic data · Precarious test, 50% overlap | miss rate: 48.45% -> 42.47% | Helped |
| Add adversarial imposters during adaptation (ablation) | RPN+ pretrained on synthetic data and adapted on T · Precarious test, 50% overlap | miss rate: 48.45% -> 45.97% | Helped |
| Fine-tune on real target images after adaptation on T ∪ I (ablation) | RPN+ pretrained on synthetic data · Precarious test, 50% overlap | miss rate: 45.97% -> 42.47% | Helped |
| Add adversarial imposters and final real-target fine-tuning (ablation) | RPN+ pretrained on synthetic data · Precarious test, 70% overlap | miss rate: 77.14% -> 73.70% | Helped |
| Fine-tune on real target images after adaptation on T ∪ I (ablation) | RPN+ pretrained on synthetic data · Precarious test, 70% overlap | miss rate: 74.94% -> 73.70% | Helped |
| Use synthetic data only instead of real-target adaptation (ablation) | RPN+ · Precarious test, 50% overlap | miss rate: 83.49% vs 42.47% | Hurt |
| Use target data only instead of synthetic pretraining and imposter adaptation (ablation) | RPN+ · Precarious test, 50% overlap | miss rate: 72.39% vs 42.47% | Hurt |
| Set the imposter set size roughly equal to the 500 target training samples (ablation) | Adversarial imposter training pipeline · Precarious test | miss rate | Helped |
| Use a better-trained real-versus-synthetic discriminator to select imposters (ablation) | Adversarial imposter training pipeline · Precarious test | miss rate | Helped |
Closest papers by meaning
- Community Forensics: Using Thousands of Generators to Train Fake Image Detectors (CVPR 2025 · similarity 0.86)
- Learning to Generate Diverse Pedestrian Movements from Web Videos with Noisy Labels (ICLR 2025 · similarity 0.85)
- Pedestrian Trajectory Prediction with Missing Data: Datasets, Imputation, and Benchmarking (NeurIPS 2024 · similarity 0.84)
- Adversarial Attacks on Event-Based Pedestrian Detectors: A Physical Approach (AAAI 2025 · similarity 0.84)
- Self-Supervised Likelihood Estimation with Energy Guidance for Anomaly Segmentation in Urban Scenes (AAAI 2024 · similarity 0.84)
- Multi-View Pedestrian Occupancy Prediction with a Novel Synthetic Dataset (AAAI 2025 · 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.