Learning to Point and Count
This addresses the what-and-where deadlock in object detection for researchers, though it appears incremental as it builds on existing detection frameworks.
The paper tackles the point-and-count problem to localize and count objects simultaneously, proposing two alternative approaches that pivot on solving 'what' or 'where' first, and evaluates them on a dataset with multiple instances of the same class to demonstrate potentials and synergies.
This paper proposes the problem of point-and-count as a test case to break the what-and-where deadlock. Different from the traditional detection problem, the goal is to discover key salient points as a way to localize and count the number of objects simultaneously. We propose two alternatives, one that counts first and then point, and another that works the other way around. Fundamentally, they pivot around whether we solve "what" or "where" first. We evaluate their performance on dataset that contains multiple instances of the same class, demonstrating the potentials and their synergies. The experiences derive a few important insights that explains why this is a much harder problem than classification, including strong data bias and the inability to deal with object scales robustly in state-of-art convolutional neural networks.