Scene-specific crowd counting using synthetic training images. (April 2022)

• Record Type: Journal Article
• Authors: Delussu, Rita; Putzu, Lorenzo; Fumera, Giorgio
• Is Part Of: Pattern recognition. Volume 124(2022)
• Publication Date: 2022-04
• Subjects: Crowd counting -- Scene-specific settings -- Synthetic training images; Pattern perception -- Periodicals; Perception des structures -- Périodiques; Patroonherkenning; 006.4
• DOI: 10.1016/j.patcog.2021.108484
• Languages: English
• ISSNs: 0031-3203

Abstract

Highlights: We propose a method for learning scene-specific crowd counting models when neither labelled nor representative unlabelled images of the target scene are available. Our method consists of generating synthetic images with the same background, perspective and scale as the target scene. Our method can be applied to any regression-based crowd counting method. Experiments on benchmark data sets show that our method can outperform the alternative cross-scene approach for both state-of-the-art CNN-based methods and early regression-based ones. Abstract: Crowd counting is a computer vision task on which considerable progress has recently been made thanks to convolutional neural networks. However, it remains a challenging task even in scene-specific settings, in real-world application scenarios where no representative images of the target scene are available, not even unlabelled, for training or fine-tuning a crowd counting model. Inspired by previous work in other computer vision tasks, we propose a simple but effective solution for the above application scenario, which consists of automatically building a scene-specific training set of synthetic images. Our solution does not require from end-users any manual annotation effort nor the collection of representative images of the target scene. Extensive experiments on several benchmark data sets show that the proposed solution can improve the effectiveness of existing crowd counting methods.