Research map / Detection and segmentation / Open-set and few-shot detection
Modeling Visual Context is Key to Augmenting Object Detection Datasets
Nikita Dvornik, Julien Mairal, Cordelia Schmid
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 context-driven copy-paste augmentation | BlitzNet single-category · VOC07-test (20 classes) | AP (%): 58.0 -> 62.0 | Improved |
| add random-placement copy-paste augmentation (ablation) | BlitzNet single-category · VOC07-test (20 classes) | AP (%): 58.0 -> 56.9 | Hurt |
| add context model to guide placement of pasted objects (ablation) | Copy-paste augmentation · VOC07-test first five classes | AP (%): 48.3 -> 57.5 | Improved |
| add context-guided placement versus removing context (ablation) | Context-DA · VOC07-test first five classes | AP (%): 33.9 -> 57.5 | Improved |
| use context-driven copy-paste instead of enlarge and re-blend augmentation (ablation) | Data augmentation · VOC07-test first five classes | AP (%): 51.6 -> 57.5 | Improved |
| add enlarge and re-blend augmentation to the base (ablation) | Base-DA · VOC07-test first five classes | AP (%): 48.7 -> 51.6 | Improved |
| add Context-DA with 50% of training data | BlitzNet · VOC07-test first five classes | AP (%): 42.2 -> 48.8 | Improved |
| add Context-DA with 25% of training data | BlitzNet · VOC07-test first five classes | AP (%): 32.5 -> 38.3 | Improved |
| add Context-DA to a multiple-category detector | BlitzNet multi-category · VOC07-test (20 classes) | AP (%): 64.6 -> 65.9 | Improved |
| add Context-DA for class chair | BlitzNet single-category · VOC07-test chair | AP (%): 40.4 -> 39.7 | Hurt |
Closest papers by meaning
- On the Importance of Visual Context for Data Augmentation in Scene Understanding (TPAMI 2019 · similarity 0.95)
- Learning Data Augmentation Strategies for Object Detection (ECCV 2020 · similarity 0.90)
- The Devil is in the Object Boundary: Towards Annotation-free Instance Segmentation using Foundation Models (ICLR 2024 · similarity 0.87)
- ContextShift: A Controlled Benchmark for Context Dependence in Object Detection (NeurIPS 2026 · similarity 0.87)
- Training-Free Open-Ended Object Detection and Segmentation via Attention as Prompts (NeurIPS 2024 · similarity 0.86)
- Multi-Perspective Data Augmentation for Few-shot Object Detection (ICLR 2025 · similarity 0.86)
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.
