Training Region-Based Object Detectors with Online Hard Example Mining

By Abhinav Shrivastava, Abhinav Kumar Gupta, Ross B. Girshick. 2016

2,741 citations recorded by the source corpus.

The field of object detection has made significant advances riding on the wave of region-based ConvNets, but their training procedure still includes many heuristics and hyperparameters that are costly to tune. We present a simple yet surprisingly effective online hard example mining (OHEM) algorithm for training region-based ConvNet detectors. Our motivation is the same as it has always been - detection datasets contain an overwhelming number of easy examples and a small number of hard examples. Automatic selection of these hard examples can make training more effective and efficient. OHEM is a simple and intuitive algorithm that eliminates several heuristics and hyperparameters in common use. But more importantly, it yields consistent and significant boosts in detection performance on benchmarks like PASCAL VOC 2007 and 2012. Its effectiveness increases as datasets become larger and more difficult, as demonstrated by the results on the MS COCO dataset. Moreover, combined with complementary advances in the field, OHEM leads to state-of-the-art results of 78.9% and 76.3% mAP on PASCAL VOC 2007 and 2012 respectively.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition

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. Trevor Darrell
2. Dacheng Tao
3. Alan L. Yuille
4. Jianfeng Gao
5. Mohit Bansal
6. Quoc V. Le
7. Luc Van Gool
8. Bernt Schiele

In Computer Vision and Pattern Recognition

1. Luc Van Gool
2. Dacheng Tao
3. Xiaogang Wang
4. Ming-Hsuan Yang
5. Trevor Darrell
6. Alan L. Yuille
7. Yi Yang
8. Chunhua Shen

Citations in the corpus

• Learning Dense Representations for Entity Retrieval (2019, methodology): “Negative sampling strategies have been employed for many models and applications, e.g. Shrivastava et al. (2016).”
• Relation Networks for Object Detection (2017, methodology): “Online hard example mining (OHEM) [40] is adopted.”
• Feature Pyramid Networks for Object Detection (2016, methodology): “Our pyramid representation greatly improves RPN’s robustness to object scale variation.”
• Cascade R-CNN: High Quality Object Detection and Instance Segmentation (2019, methodology): “Online hard negative mining [50] was not used.”
• Fully Convolutional Instance-Aware Semantic Segmentation (2016, methodology): “Specifically, we apply online hard example mining (OHEM) [38].”
• Beyond Skip Connections: Top-Down Modulation for Object Detection (2016, background): “Convolutional neural networks (ConvNets) have revolutionized the field of object detection [4, 14, 16, 17, 41, 44, 48].”
• Speed/Accuracy Trade-Offs for Modern Convolutional Object Detectors (2016, background): “Since appearing in 2015, Faster R-CNN has been particularly influential, and has led to a number of follow-up works [3, 30, 29, 40, 12, 5] (including SSD and R-FCN).”
• R-FCN: Object Detection via Region-based Fully Convolutional Networks (2016, methodology): “Following R-CNN [8], we adopt the popular two-stage object detection strategy [8, 9, 6, 7, 19, 1, 23] that consists of: (i) region proposal, and (ii) region classification.”
• iShape: A First Step Towards Irregular Shape Instance Segmentation (2021, positive): “Since affinity ground truth is extremely imbalanced, an OHEM [35] loss is calculated between the predicted affinity map and the affinity ground truth to effectively alleviate the problem.”
• Focal Loss for Dense Object Detection (2017, background): “We also implement a variant of OHEM used in SSD [21]: after applying nms to all examples, the minibatch is constructed to enforce a 1:3 ratio between positives and negatives to help ensure each minibatch has enough positives.”
• CoupleNet: Coupling Global Structure with Local Parts for Object Detection (2017, methodology): “For a fair comparison, we first re-implemented Faster RCNN [12] using ResNet-101 and online hard example mining (OHEM) [22], which achieves a mAP of 79.”
• Active Bias: Training More Accurate Neural Networks by Emphasizing High Variance Samples (2017, methodology): “This strategy is sometimes referred to as hard example mining [44].”
• MMDetection: Open MMLab Detection Toolbox and Benchmark (2019, background): “• OHEM [29]: an online sampling method that mines hard samples for training, proposed in 2016.”