Few-shot Object Counting and Detection

Few-shot Object Counting and Detection aims to count and detect instances of objects of the same class as given exemplars within a single image, using only up to three exemplars from the original FSC147. This task is especially challenging but crucial when large datasets are unavailable. Methods for this task must output a single set of bounding boxes, from which the final count is derived based on the number of detected boxes.