ePrints@IIScePrints@IISc Home | About | Browse | Latest Additions | Advanced Search | Contact | Help

Completely Self-supervised Crowd Counting via Distribution Matching

Babu Sam, D and Agarwalla, A and Joseph, J and Sindagi, VA and Babu, RV and Patel, VM (2022) Completely Self-supervised Crowd Counting via Distribution Matching. In: 17th European Conference on Computer Vision, ECCV 2022, 23 - 27 October 2022, Tel Aviv, pp. 186-204.

ECCV_2022.pdf - Published Version

Download (11MB) | Preview
Official URL: https://doi.org/10.1007/978-3-031-19821-2_11


Dense crowd counting is a challenging task that demands millions of head annotations for training models. Though existing self-supervised approaches could learn good representations, they require some labeled data to map these features to the end task of density estimation. We mitigate this issue with the proposed paradigm of complete self-supervision, which does not need even a single labeled image. The only input required to train, apart from a large set of unlabeled crowd images, is the approximate upper limit of the crowd count for the given dataset. Our method dwells on the idea that natural crowds follow a power law distribution, which could be leveraged to yield error signals for backpropagation. A density regressor is first pretrained with self-supervision and then the distribution of predictions is matched to the prior. Experiments show that this results in effective learning of crowd features and delivers significant counting performance. © 2022, The Author(s), under exclusive license to Springer Nature Switzerland AG.

Item Type: Conference Paper
Publication: Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Publisher: Springer Science and Business Media Deutschland GmbH
Additional Information: The copyright for this article belongs to the Author(S).
Keywords: Backpropagation; Computer vision, Crowd counting; Density estimation; Distribution matching; Head annotations; Labeled data; Labeled images; Learn+; Self-supervision; Training model; Upper limits, Large dataset
Department/Centre: Division of Electrical Sciences > Electrical Communication Engineering
Date Deposited: 09 Jan 2023 09:02
Last Modified: 09 Jan 2023 09:02
URI: https://eprints.iisc.ac.in/id/eprint/78932

Actions (login required)

View Item View Item