YOLOv4: Optimal Speed and Accuracy of Object Detection

TLDR

CNN accuracy can be improved by many features, some of which are model‑ or dataset‑specific while others, like batch‑normalization and residual connections, are broadly applicable. The study aims to test combinations of universal CNN features—Weighted‑Residual‑Connections, Cross‑Stage‑Partial connections, Cross mini‑Batch Normalization, Self‑adversarial training, and Mish activation—on large datasets and provide theoretical justification. By integrating WRC, CSP, CmBN, SAT, Mish activation, Mosaic augmentation, DropBlock, and CIoU loss, the authors achieve 43.5 % AP (65.7 % AP50) on MS COCO at ~65 FPS on a Tesla V100. Source code is available at https://github.com/AlexeyAB/darknet.

Abstract

There are a huge number of features which are said to improve Convolutional Neural Network (CNN) accuracy. Practical testing of combinations of such features on large datasets, and theoretical justification of the result, is required. Some features operate on certain models exclusively and for certain problems exclusively, or only for small-scale datasets; while some features, such as batch-normalization and residual-connections, are applicable to the majority of models, tasks, and datasets. We assume that such universal features include Weighted-Residual-Connections (WRC), Cross-Stage-Partial-connections (CSP), Cross mini-Batch Normalization (CmBN), Self-adversarial-training (SAT) and Mish-activation. We use new features: WRC, CSP, CmBN, SAT, Mish activation, Mosaic data augmentation, CmBN, DropBlock regularization, and CIoU loss, and combine some of them to achieve state-of-the-art results: 43.5% AP (65.7% AP50) for the MS COCO dataset at a realtime speed of ~65 FPS on Tesla V100. Source code is at https://github.com/AlexeyAB/darknet