Prototype distance ratio sampling for generalised few shot object detection

• By comparing a wide range of few-shot learning methods, the research in this paper shows that gradient correction methods can harm Generalised Few shot Object Detection (GFSOD) performance unless the gradients are averaged appropriately. • A selection of different few shot learning configurations were evaluated, including changing the base sampling limit and sample balancing, demonstrating that these approaches can improve base and novel Average Precision (AP). • The experiments show that selecting base samples depending on their absolute distance to their respective prototype yields a base AP improvement, and using the ratio of distances to the correct prototype yields further improvement specifically on the widely used benchmark COCO dataset. Few-Shot Learning has emerged as a topic that maximises DNN performance based on very few samples. In Generalised Few-Shot Learning, a model has to learn new few-shot classes while recalling earlier large-scale training classes. Learning the new classes leads to a drop in performance on the base ones. In this work, we identify and explore the parallels between Generalised Few-Shot Object Detection (G-FSOD) and Continual Learning (CL), focusing on two areas in particular: gradient manipulation methods and sampling strategies. Through extensive experimentation we demonstrate that gradient manipulation methods appear to be no better than existing techniques and do not improve performance, but actually harm performance unless the gradients are averaged. Our investigations into sampling strategies consider a number of aspects: the impact of removing the base limit and the effectiveness of different distance measures (with respect to a class prototype) for sample selection. Our experiments into these aspects reveal illuminating insights into their impact on Average Precision on the COCO and VOC datasets. Consequently, we suggest that G-FSOD research focus on the replay aspect and investigate other sampling strategies.