Research map / Detection and segmentation / Open-set and few-shot detection
Learning Object Context for Dense Captioning
Read the paperCheck 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 |
|---|---|---|---|
| use learned object context as guidance (COCG) instead of JIVC whole-image context | dense captioning model · VG-COCO | mAP (%): 7.85 -> 8.90 | Helped |
| decode learned object context with an LSTM (COCD) instead of whole-image context | JIVC · VG-COCO | mAP (%): 7.85 -> 7.92 | No clear effect |
| replace context LSTM with max pooling over objects (ablation) | COCG · VG-COCO | mAP (%): 8.90 -> 7.86 | Hurt |
| add region-specific learned context instead of whole image context as guidance (ablation) | COCG · VG-COCO | mAP (%): 7.81 -> 8.90 | Helped |
| add location and size features to object representation (ablation) | COCG · VG-COCO | mAP (%): 8.76 -> 8.90 | Helped |
| use ground truth objects instead of detected objects (ablation) | COCG · VG-COCO | mAP (%): 8.90 -> 9.79 | Helped |
| sort objects by area instead of location (ablation) | COCG · VG-COCO | mAP (%): 8.90 -> 8.74 | Hurt |
| sort objects by confidence instead of location (ablation) | COCG · VG-COCO | mAP (%): 8.90 -> 8.89 | No clear effect |
| use COCG on VG V1.0 | JIVC · VG V1.0 | mAP (%): 9.31 -> 9.82 | Helped |
| use COCG on VG V1.2 | JIVC · VG V1.2 | mAP (%): 9.96 -> 10.39 | Helped |
| use COCD instead of JIVC | JIVC · VG V1.2 | mAP (%): 9.96 -> 9.75 | Hurt |
Closest papers by meaning
- Dense Captioning With Joint Inference and Visual Context (CVPR 2017 · similarity 0.91)
- Beyond a Pre-Trained Object Detector: Cross-Modal Textual and Visual Context for Image Captioning (CVPR 2022 · similarity 0.90)
- CALVIN: Improved Contextual Video Captioning via Instruction Tuning (NeurIPS 2024 · similarity 0.87)
- Dense Video Object Captioning from Disjoint Supervision (ICLR 2025 · similarity 0.87)
- FlexCap: Describe Anything in Images in Controllable Detail (NeurIPS 2024 · similarity 0.87)
- Relational Distant Supervision for Image Captioning without Image-Text Pairs (AAAI 2024 · 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.