文献类型：期刊文献

英文题名：TranSDet: Toward Effective Transfer Learning for Small-Object Detection

作者：Xu, Xinkai[1,2,3];Zhang, Hailan[1];Ma, Yan[2,3];Liu, Kang[1];Bao, Hong[2,3];Qian, Xu[1]

第一作者：Xu, Xinkai

通讯作者：Ma, Y[1];Ma, Y[2]

机构：[1]China Univ Min & Technol Beijing, Sch Artificial Intelligence, Beijing 100083, Peoples R China;[2]Beijing Union Univ, Beijing Key Lab Informat Serv Engn, Beijing 100101, Peoples R China;[3]Beijing Union Univ, Coll Robot, Beijing 100027, Peoples R China

第一机构：China Univ Min & Technol Beijing, Sch Artificial Intelligence, Beijing 100083, Peoples R China

通讯机构：[1]corresponding author), Beijing Union Univ, Beijing Key Lab Informat Serv Engn, Beijing 100101, Peoples R China;[2]corresponding author), Beijing Union Univ, Coll Robot, Beijing 100027, Peoples R China.|[1141739]北京联合大学机器人学院;[11417]北京联合大学;[11417103]北京联合大学北京市信息服务工程重点实验室;

年份：2023

卷号：15

期号：14

外文期刊名：REMOTE SENSING

收录：;EI(收录号：20233114470571);Scopus(收录号：2-s2.0-85166230446);WOS:【SCI-EXPANDED(收录号:WOS:001036488700001)】；

语种：英文

外文关键词：object detection; transfer learning; dynamic resolution adaptation; small-object detection

摘要：Small-object detection is a challenging task in computer vision due to the limited training samples and low-quality images. Transfer learning, which transfers the knowledge learned from a large dataset to a small dataset, is a popular method for improving performance on limited data. However, we empirically find that due to the dataset discrepancy, directly transferring the model trained on a general object dataset to small-object datasets obtains inferior performance. In this paper, we propose TranSDet, a novel approach for effective transfer learning for small-object detection. Our method adapts a model trained on a general dataset to a small-object-friendly model by augmenting the training images with diverse smaller resolutions. A dynamic resolution adaptation scheme is employed to ensure consistent performance on various sizes of objects using meta-learning. Additionally, the proposed method introduces two network components, an FPN with shifted feature aggregation and an anchor relation module, which are compatible with transfer learning and effectively improve small-object detection performance. Extensive experiments on the TT100K, BUUISE-MO-Lite, and COCO datasets demonstrate that TranSDet achieves significant improvements compared to existing methods. For example, on the TT100K dataset, TranSDet outperforms the state-of-the-art method by 8.0% in terms of the mean average precision (mAP) for small-object detection. On the BUUISE-MO-Lite dataset, TranSDet improves the detection accuracy of RetinaNet and YOLOv3 by 32.2% and 12.8%, respectively.