Shot Object Detection
Few-shot object detection (FSOD) aims to train object detectors that can accurately identify and locate objects with minimal labeled training data, addressing the limitations of traditional methods requiring massive datasets. Current research focuses on improving model efficiency and accuracy through techniques like meta-learning, data augmentation (including synthetic data generation), and the integration of vision-language models, often within two-stage or one-stage detector architectures. These advancements are significant because they enable faster and more cost-effective development of object detection systems, particularly for applications involving rare or newly emerging object categories in fields like remote sensing and robotics.
Papers
Multi-Faceted Distillation of Base-Novel Commonality for Few-shot Object Detection
Shuang Wu, Wenjie Pei, Dianwen Mei, Fanglin Chen, Jiandong Tian, Guangming Lu
Rethinking Few-Shot Object Detection on a Multi-Domain Benchmark
Kibok Lee, Hao Yang, Satyaki Chakraborty, Zhaowei Cai, Gurumurthy Swaminathan, Avinash Ravichandran, Onkar Dabeer
Few-shot Object Counting and Detection
Thanh Nguyen, Chau Pham, Khoi Nguyen, Minh Hoai