This paper presents a detection-aware pre-training (DAP) approach, which leverages only weakly-labeled classification-style datasets (e.g., ImageNet) for pre-training, but is specifically tailored to benefit object detection tasks. In contrast to the widely used image classification-based pre-training (e.g., on ImageNet), which does not include any location-related training tasks, we transform a classification dataset into a detection dataset through a weakly supervised object localization method based on Class Activation Maps to directly pre-train a detector, making the pre-trained model location-aware and capable of predicting bounding boxes. We show that DAP can outperform the traditional classification pre-training in terms of both sample efficiency and convergence speed in downstream detection tasks including VOC and COCO. In particular, DAP boosts the detection accuracy by a large margin when the number of examples in the downstream task is small.
@article{arxiv.2103.16651,
title = {DAP: Detection-Aware Pre-training with Weak Supervision},
author = {Yuanyi Zhong and Jianfeng Wang and Lijuan Wang and Jian Peng and Yu-Xiong Wang and Lei Zhang},
journal= {arXiv preprint arXiv:2103.16651},
year = {2021}
}