StarNet: towards weakly supervised few-shot detection and explainable few-shot classification



Published on



AAAI Computer Vision

Few-shot detection and classification have advanced significantly in recent years. Yet, detection approaches require strong annotation (bounding boxes) both for pre-training and for adaptation to novel classes, and classification approaches rarely provide localization of objects in the scene. In this paper, we introduce StarNet – a few-shot model featuring an end-to-end differentiable non-parametric star-model detection and classification head. Through this head, the backbone is meta-trained using only image-level labels to produce good features for jointly localizing and classifying previously unseen categories of few-shot test tasks using a star-model that geometrically matches between the query and support images (to find corresponding object instances). Being a few-shot detector, StarNet does not require any bounding box annotations, neither during pre-training nor for novel classes adaptation. It can thus be applied to the previously unexplored and challenging task of Weakly Supervised Few-Shot Object Detection (WS-FSOD), where it attains significant improvements over the baselines. In addition, StarNet shows significant gains on few-shot classification benchmarks that are less cropped around the objects (where object localization is key).

This paper has been published at AAAI 2020

Please cite our work using the BibTeX below.

      title={StarNet: towards Weakly Supervised Few-Shot Object Detection}, 
      author={Leonid Karlinsky and Joseph Shtok and Amit Alfassy and Moshe Lichtenstein and Sivan Harary and Eli Schwartz and Sivan Doveh and Prasanna Sattigeri and Rogerio Feris and Alexander Bronstein and Raja Giryes},
Close Modal