Add to Chrome
Add to Firefox
Add to EdgeComparison Network for One-Shot Conditional Object Detection
Apr 04, 2019
The current advances in object detection depend on large-scale datasets to get good performance. However, there may not always be sufficient samples in many scenarios, which leads to the research on few-shot detection as well as its extreme variation one-shot detection. In this paper, the one-shot detection has been formulated as a conditional probability problem. With this insight, a novel one-shot conditional object detection (OSCD) framework, referred as Comparison Network (ComparisonNet), has been proposed. Specifically, query and target image features are extracted through a Siamese network as mapped metrics of marginal probabilities. A two-stage detector for OSCD is introduced to compare the extracted query and target features with the learnable metric to approach the optimized non-linear conditional probability. Once trained, ComparisonNet can detect objects of both seen and unseen classes without further training, which also has the advantages including class-agnostic, training-free for unseen classes, and without catastrophic forgetting. Experiments show that the proposed approach achieves state-of-the-art performance on the proposed datasets of Fashion-MNIST and PASCAL VOC.
