In this work, we propose a novel approach that allows for the end-to-end learning of multi-instance point detection with inherent sub-pixel precision capabilities. To infer unambiguous localization estimates, our model relies on three components: the continuous prediction capabilities of offset-regression-based models, the finer-grained spatial learning ability of a novel continuous heatmap matching loss function introduced to that effect, and the prediction sparsity ability of count-based regularization. We demonstrate strong sub-pixel localization accuracy on single molecule localization microscopy and checkerboard corner detection, and improved sub-frame event detection performance in sport videos.
CITATION STYLE
Schroeter, J., Tuytelaars, T., Sidorov, K., & Marshall, D. (2021). Learning Multi-instance Sub-pixel Point Localization. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 12626 LNCS, pp. 669–686). Springer Science and Business Media Deutschland GmbH. https://doi.org/10.1007/978-3-030-69541-5_40
Mendeley helps you to discover research relevant for your work.