Research map / Detection and segmentation / CNN features
DetCo: Unsupervised Contrastive Learning for Object Detection
Enze Xie, Jian Ding, Wenhai Wang, Xiaohang Zhan, Hang Xu, Peize Sun, Zhenguo Li, Ping Luo
This paper tried a common change: Add a contrastive loss
add global and local contrastive learning · ImageNet linear classification (100ep) · top-1 accuracy: 64.3 -> 67.1 Improved
What 1,317 accepted papers report for this change
- 1,253 improved
- 22 no clear effect
- 16 mixed
- 26 hurt
- Beyond Looking Up, Try Looking Around: Harmonizing Global Structure and Local Consistency in Optimal Transport for Short Text Clustering ICML 2026 · add contrastive learning · SearchSnippets · ACC: 56.65 -> 80.12Improved
- SleepLM: Natural-Language Intelligence for Human Sleep ICML 2026 · add contrastive objective to captioning (CoCa vs Cap) · sleep event IoU · IoU: 29.2 -> 27.4Hurt
- LIMSSR: LLM-Driven Sequence-to-Score Reasoning under Training-Time Incomplete Multimodal Observations ICML 2026 · add contrastive loss L_con · FS1000 average · Spearman correlation: 0.759 -> 0.789Improved
See this change in detailMatched by the wording of each paper's own result, read by AI from the full text. Papers report gains more often than failures, so this shows what gets published, not the chance a change works.
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 multi-level supervision (ablation) | MoCo v2 · VOC07+12 detection (100ep) | mAP: 56.3 -> 57.0 | Improved |
| add multi-level supervision (ablation) | MoCo v2 · ImageNet linear classification (100ep) | top-1 accuracy: 64.3 -> 63.2 | Hurt |
| add multi-level supervision (ablation) | MoCo v2 · VOC07 SVM Res2 feature | accuracy: 47.1 -> 50.9 | Improved |
| add multi-level supervision (ablation) | MoCo v2 · VOC07 SVM Res5 feature | accuracy: 82.1 -> 81.8 | Hurt |
| add global and local contrastive learning (ablation) | MoCo v2 · ImageNet linear classification (100ep) | top-1 accuracy: 64.3 -> 67.1 | Improved |
| add global and local contrastive learning (ablation) | MoCo v2 · VOC07+12 detection (100ep) | mAP: 56.3 -> 56.8 | Improved |
| add global and local contrastive learning on top of multi-level supervision (ablation) | MoCo v2 + MLS · VOC07+12 detection (100ep) | mAP: 57.0 -> 57.4 | Improved |
| add global and local contrastive learning on top of multi-level supervision (ablation) | MoCo v2 + MLS · ImageNet linear classification (100ep) | top-1 accuracy: 63.2 -> 66.6 | Improved |
| DetCo pretraining instead of MoCo v2 under 12k-iteration fine-tuning | Mask R-CNN R50-C4 · COCO 12k | AP50^bb: 45.3 -> 49.1 | Improved |
| pretrain DetCo 800 epochs instead of 200 (ablation) | DetCo · VOC07+12 detection | AP: 57.8 -> 58.2 | Improved |
| DetCo with 100 epochs vs MoCo v2 with 800 epochs | DetCo · VOC07+12 detection | AP: 57.4 -> 57.4 | No clear effect |
| use DetCo pretraining instead of supervised pretraining | Sparse R-CNN · COCO | AP: 45.0 -> 46.5 | Improved |
Later papers that compared with this method
1 later accepted paper reports DetCo as a baseline in their results tables; 0 of them report a better number on most of the results they share with it.
- Deeply Unsupervised Patch Re-Identification for Pre-Training Object Detectors (TPAMI 2022 · better on 0 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
- Sparse DETR: Efficient End-to-End Object Detection with Learnable Sparsity (ICLR 2022 · similarity 0.90)
- Incremental-DETR: Incremental Few-Shot Object Detection via Self-Supervised Learning (AAAI 2023 · similarity 0.90)
- End-to-End Object Detection with Transformers (ECCV 2020 · similarity 0.90)
- Proposal-Contrastive Pretraining for Object Detection from Fewer Data (ICLR 2023 · similarity 0.89)
- Meta-DETR: Image-Level Few-Shot Detection with Inter-Class Correlation Exploitation (TPAMI 2022 · similarity 0.89)
- MS-DETR: Efficient DETR Training with Mixed Supervision (CVPR 2024 · similarity 0.88)
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.
