Research map / Detection and segmentation / Open-set and few-shot detection
COBE: Contextualized Object Embeddings from Narrated Instructional Video
Gedas Bertasius, Lorenzo Torresani
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 |
|---|---|---|---|
| replace CTRL with RoBERTa as the language model (ablation) | COBE · HowTo100M_BB test | mAP: 18.4 -> 9.4 | Hurt |
| replace CTRL with T5 (ablation) | COBE · HowTo100M_BB test | mAP: 18.4 -> 14.1 | Hurt |
| replace CTRL with BERT (ablation) | COBE · HowTo100M_BB test | mAP: 18.4 -> 16.2 | Hurt |
| replace CTRL with XLM (ablation) | COBE · HowTo100M_BB test | mAP: 18.4 -> 18.0 | Hurt |
| replace CTRL with Word2Vec (non-contextual) (ablation) | COBE · HowTo100M_BB test | mAP: 18.4 -> 16.7 | Hurt |
| reduce negatives per positive from 2048 to 512 (ablation) | COBE · HowTo100M_BB test | mAP: 18.4 -> 15.5 | Hurt |
| reduce negatives per positive from 2048 to 128 (ablation) | COBE · HowTo100M_BB test | mAP: 18.4 -> 13.6 | Hurt |
| pretrain on HowTo100M_BB then finetune on EK vs training only on EK | COBE · EPIC-Kitchens detection | mAP: 12.7 -> 22.6 | Improved |
| COBE instead of Faster R-CNN for object detection | COBE · EPIC-Kitchens detection | mAP: 14.0 -> 15.4 | Improved |
| COBE instead of Faster R-CNN + 3D ResNet-101 (Kinetics) | COBE · EK_K (noun, noun) | mAP: 7.7 -> 32.4 | Improved |
| COBE instead of Faster R-CNN + 2D ResNet-101 (ImageNet) | COBE · EK_I (noun, noun) | mAP: 13.2 -> 26.0 | Improved |
| COBE instead of Faster R-CNN + S3D | COBE · EK_H (noun, verb) | mAP: 14.1 -> 24.7 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- Faster R-CNN — Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks (TPAMI 2016) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Learning Object State Changes in Videos: An Open-World Perspective (CVPR 2024 · similarity 0.88)
- Benefit From Seen: Enhancing Open-Vocabulary Object Detection by Bridging Visual and Textual Co-Occurrence Knowledge (ICCV 2025 · similarity 0.88)
- The Devil is in the Fine-Grained Details: Evaluating Open-Vocabulary Object Detectors for Fine-Grained Understanding (CVPR 2024 · similarity 0.87)
- Open-vocabulary Object Detection via Vision and Language Knowledge Distillation (ICLR 2022 · similarity 0.87)
- ORIC: Benchmarking Object Recognition under Contextual Incongruity in Large Vision-Language Models (CVPR 2026 · similarity 0.87)
- DoGA: Enhancing Grounded Object Detection via Grouped Pre-Training with Attributes (AAAI 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.
