DetCo: Unsupervised Contrastive Learning for Object Detection

By Enze Xie, Jian Ding, Wenhai Wang, Xiaohang Zhan, Hang Xu, Zhenguo Li, Ping Luo. IEEE International Conference on Computer Vision, 2021

369 citations recorded by the source corpus.

We present DetCo, a simple yet effective self-supervised approach for object detection. Unsupervised pre-training methods have been recently designed for object detection, but they are usually deficient in image classification, or the opposite. Unlike them, DetCo transfers well on downstream instance-level dense prediction tasks, while maintaining competitive image-level classification accuracy. The advantages are derived from (1) multi-level supervision to intermediate representations, (2) contrastive learning between global image and local patches. These two designs facilitate discriminative and consistent global and local representation at each level of feature pyramid, improving detection and classification, simultaneously.Extensive experiments on VOC, COCO, Cityscapes, and ImageNet demonstrate that DetCo not only outperforms recent methods on a series of 2D and 3D instance-level detection tasks, but also competitive on image classification. For example, on ImageNet classification, DetCo is 6.9% and 5.0% top-1 accuracy better than InsLoc and DenseCL, which are two contemporary works designed for object detection. Moreover, on COCO detection, DetCo is 6.9 AP better than SwAV with Mask R-CNN C4. Notably, DetCo largely boosts up Sparse R-CNN, a recent strong detector, from 45.0 AP to 46.5 AP (+1.5 AP), establishing a new SOTA on COCO.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/77f188254b81fd9d3789404a6ea02591d4173123

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Dacheng Tao
2. Xiaodan Liang
3. Jiashi Feng
4. Zhenguo Li
5. Qi Tian
6. Caiming Xiong
7. Hang Xu
8. Jing Zhang

In Computer Vision and Pattern Recognition

1. Dacheng Tao
2. Luc Van Gool
3. Jiashi Feng
4. Yi Yang
5. Alan L. Yuille
6. Qi Tian
7. Xiaogang Wang
8. Ming-Hsuan Yang

Citations in the corpus

• SODA10M: Towards Large-Scale Object Detection Benchmark for Autonomous Driving (2021, methodology): “We follow the default settings in OpenSelfSup1 to train six state-of-the-art standard self-supervised learning methods, including MoCo-v1 [21], MoCo-v2 [5], SimCLR [4], SwAV [3], DetCo [58], DenseCL [56], and evaluate their performance by fine-tuning the pre-trained models on the SODA10M labeled dat”
• Open Vocabulary Object Detection with Pseudo Bounding-Box Labels (2021, background): “Object detection [11, 19, 20, 27] is a core task in computer vision that has considerably advanced with the adoption of deep learning and continues to attract significant research effort [26, 31, 33].”
• Dual Temperature Helps Contrastive Learning Without Many Negative Samples: Towards Understanding and Simplifying MoCo (2022, background): “CL remains as a mainstream framework for SSL and has also been extensively studied in other filed applications [1,13,16,27,38,47,52,55,56,59,65].”
• Federated Learning from Pre-Trained Models: A Contrastive Learning Approach (2022, background): “Contrastive learning has been widely applied to self-supervised learning scenarios in recent years, achieving state-of-the-art performance in the unsupervised training of deep image models [33, 34] and graph models [35, 36, 37].”
• Patch-level Representation Learning for Self-supervised Vision Transformers (2022, methodology): “We consider a variety of existing SSL methods developed for ResNet [20] and ViT [35] architectures: (a) self-supervised ResNets: MoCo-v2 [7], SwAV [3], DenseCL [39], ReSim [43], and DetCo [44]; and (b) selfsupervised ViTs: DINO [4], MoCo-v3 [9], and MoBY [46].”
• How Well Self-Supervised Pre-Training Performs with Streaming Data? (2021, background): “Recently, instance discrimination has become the most popular pre-text task for SSL, which motivates various contrastive SSL methods [53, 7, 19, 9, 37, 54, 59].”
• SPot-the-Difference Self-Supervised Pre-training for Anomaly Detection and Segmentation (2022, background): “Moreover, to promote spatial details of representations for localization tasks, several approaches proposed to encourage the invariance of patch features to global augmentations [7,28,41,42], although they may not lead to local sensitivity to tiny defects.”
• A Survey on Self-Supervised Learning: Algorithms, Applications, and Future Trends (2023, background): “…distillation [188], optical flow prediction [189], vision-language navigation [190], physiological signal estimation [191], [192], image denoising [193], [194], object detection [195], [196], [197], super-resolution [198], [199], voxel prediction from 2D images [200], and ego-motion [201], [202].”
• Seed the Views: Hierarchical Semantic Alignment for Contrastive Representation Learning (2022, background): “Detco [41], DUPR [42] and OBoW [36] add the MLP heads upon intermediate layers directly, which is shown to be not suitable for CSML.”
• Joint Learning of Localized Representations from Medical Images and Reports (2021, background): “Therefore, there is a number of recent approaches that use region-level contrast[87,86,83,6,61,52], i.”