Research map / Detection and segmentation / CNN features
Seeing without Looking: Contextual Rescoring of Object Detections for AP Maximization
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 |
|---|---|---|---|
| Apply contextual confidence rescoring using all detections' confidence, class, and box location | Faster R-CNN and Cascade R-CNN · COCO val2017 | AP: 36.4 -> 37.4; 39.4 -> 39.9; 41.1 -> 41.8; 42.1 -> 42.8 | Helped |
| Train with IoU targets rather than binary targets (ablation) | Contextual rescoring · COCO val2017 | AP: 42.5 -> 42.8 | Helped |
| Include original detection confidence as an input feature (ablation) | Contextual rescoring · COCO val2017 | AP: 39.9 -> 42.8 | Helped |
| Include class and coordinates with confidence rather than confidence alone (ablation) | Contextual rescoring · COCO val2017 | AP: 42.2 -> 42.8 | Helped |
| Include bounding-box coordinates (ablation) | Contextual rescoring · COCO val2017 | AP: 42.4 -> 42.8 | Helped |
| Include predicted class (ablation) | Contextual rescoring · COCO val2017 | AP: 42.3 -> 42.8 | Helped |
| Replace an unidirectional GRU with a bidirectional GRU (ablation) | Contextual rescoring · COCO val2017 | AP: 42.6 -> 42.8 | Helped |
| Add self-attention to a unidirectional GRU (ablation) | Contextual rescoring · COCO val2017 | AP: 42.6 -> 42.7 | Helped |
| Use an LSTM instead of a GRU (ablation) | Contextual rescoring · COCO val2017 | AP: 42.6 | No clear effect |
| Use a linear layer with self-attention instead of an RNN (ablation) | Contextual rescoring · COCO val2017 | AP: 42.1 -> 42.6 | Helped |
Closest papers by meaning
- Spatially Invariant Unsupervised Object Detection with Convolutional Neural Networks (AAAI 2019 · similarity 0.90)
- ContextShift: A Controlled Benchmark for Context Dependence in Object Detection (NeurIPS 2026 · similarity 0.89)
- Exploring the Bounds of the Utility of Context for Object Detection (CVPR 2019 · similarity 0.89)
- Don't Even Look Once: Synthesizing Features for Zero-Shot Detection (CVPR 2020 · similarity 0.89)
- Unsupervised Object Detection with Theoretical Guarantees (NeurIPS 2024 · similarity 0.87)
- Believing is Seeing: Unobserved Object Detection using Generative Models (CVPR 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.