Towards Accurate One-Stage Object Detection With AP-Loss

By Kean Chen, Jianguo Li, Weiyao Lin, John See, Ji Wang, Ling-yu Duan, Zhibo Chen, Changwei He, Junni Zou. 2019

130 citations recorded by the source corpus.

One-stage object detectors are trained by optimizing classification-loss and localization-loss simultaneously, with the former suffering much from extreme foreground-background class imbalance issue due to the large number of anchors. This paper alleviates this issue by proposing a novel framework to replace the classification task in one-stage detectors with a ranking task, and adopting the Average-Precision loss (AP-loss) for the ranking problem. Due to its non-differentiability and non-convexity, the AP-loss cannot be optimized directly. For this purpose, we develop a novel optimization algorithm, which seamlessly combines the error-driven update scheme in perceptron learning and backpropagation algorithm in deep networks. We verify good convergence property of the proposed algorithm theoretically and empirically. Experimental results demonstrate notable performance improvement in state-of-the-art one-stage detectors based on AP-loss over different kinds of classification-losses on various benchmarks, without changing the network architectures.

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

In Computer Vision and Pattern Recognition

Citations in the corpus

Bridging the Gap Between Anchor-Based and Anchor-Free Detection via Adaptive Training Sample Selection (2019, methodology): “Thereafter, plenty of works are presented to boost its performance in different aspects, such as fusing context information from different layers [24, 12, 69], training from scratch [50, 73], introducing new loss function [33, 6], anchor refinement and matching [66, 67], architecture redesign [21, 2”
Improving Multispectral Pedestrian Detection by Addressing Modality Imbalance Problems (2020, background): “AP-Loss [5] and DR Loss [35] also provide ideas of designing loss function to solve the class imbalance problem.”
A Survey of Deep Learning-Based Object Detection (2019, methodology): “[88] consider design a novel ranking task to replace the conventional classification task and a newly Average-Precision”
Imbalance Problems in Object Detection: A Review (2019, methodology): “As an example, AP Loss [61] is computed considering the ranking of the individual examples, which is based on the confidence scores of all BBs.”
Prime Sample Attention in Object Detection (2019, background): “AP loss[2] and DR loss[19] introduce a new perspective that converts the classification task into a ranking task.”
Probabilistic Anchor Assignment with IoU Prediction for Object Detection (2020, background): “Since the seminal work of R-CNN [8], recent advances in object detection have shown rapid improvements with many innovative architectural designs [21,28,41,43], training objectives [3, 7, 22, 29] and post-processing schemes [2, 13, 15] with strong CNN backbones [5,11,17,19,31,32,36].”
Dice Loss for Data-imbalanced NLP Tasks (2019, methodology): “(2019) proposed a novel method called IoU-balanced sampling and Chen et al. (2019) designed a ranking model to replace the conventional classification task with an average-precision loss to alleviate the class imbalance issue.”
Enriched Feature Guided Refinement Network for Object Detection (2019, result): “6% in terms of COCO AP, compared to the best reported singlestage results [5].”