Noisy annotations such as missing annotations and location shifts often exist in crowd counting datasets due to multi-scale head sizes, high occlusion, etc. These noisy annotations severely affect the model training, especially for density map-based methods. To alleviate the negative impact of noisy annotations, we propose a novel crowd counting model with one convolution head and one transformer head, in which these two heads can supervise each other in noisy areas, called Cross-Head Supervision. The resultant model, CHS-Net, can synergize different types of inductive biases for better counting. In addition, we develop a progressive cross-head supervision learning strategy to stabilize the training process and provide more reliable supervision. Extensive experimental results on ShanghaiTech and QNRF datasets demonstrate superior performance over state-of-the-art methods. Code is available at https://github.com/RaccoonDML/CHSNet.
Monitoring and analyzing stereotypical behaviours is important for early intervention and care taking in Autism Spectrum Disorder (ASD). This paper focuses on automatically detecting stereotypical behaviours with computer vision techniques. Off-the-shelf methods tackle this task by supervised classification and activity recognition techniques. However, the unbounded types of stereotypical behaviours and the difficulty in collecting video recordings of ASD patients largely limit the feasibility of the existing supervised detection methods. As a result, we tackle these challenges from a new perspective, i.e. unsupervised video anomaly detection for stereotypical behaviours detection. The models can be trained among unlabeled videos containing only normal behaviours and unknown types of abnormal behaviours can be detected during inference. Correspondingly, we propose a Dual Stream deep model for Stereotypical Behaviours Detection, DS-SBD, based on the temporal trajectory of human poses and the repetition patterns of human actions. Extensive experiments are conducted to verify the effectiveness of our proposed method and suggest that it serves as a potential benchmark for future research.
Crowd Counting has important applications in public safety and pandemic control. A robust and practical crowd counting system has to be capable of continuously learning with the new-coming domain data in real-world scenarios instead of fitting one domain only. Off-the-shelf methods have some drawbacks to handle multiple domains. 1) The models will achieve limited performance (even drop dramatically) among old domains after training images from new domains due to the discrepancies of intrinsic data distributions from various domains, which is called catastrophic forgetting. 2) The well-trained model in a specific domain achieves imperfect performance among other unseen domains because of the domain shift. 3) It leads to linearly-increased storage overhead either mixing all the data for training or simply training dozens of separate models for different domains when new ones are available. To overcome these issues, we investigate a new task of crowd counting under the incremental domains training setting, namely, Lifelong Crowd Counting. It aims at alleviating the catastrophic forgetting and improving the generalization ability using a single model updated by the incremental domains. To be more specific, we propose a self-distillation learning framework as a benchmark~(Forget Less, Count Better, FLCB) for lifelong crowd counting, which helps the model sustainably leverage previous meaningful knowledge for better crowd counting to mitigate the forgetting when the new data arrive. Meanwhile, a new quantitative metric, normalized backward transfer~(nBwT), is developed to evaluate the forgetting degree of the model in the lifelong learning process. Extensive experimental results demonstrate the superiority of our proposed benchmark in achieving a low catastrophic forgetting degree and strong generalization ability.
Most conventional crowd counting methods utilize a fully-supervised learning framework to learn a mapping between scene images and crowd density maps. Under the circumstances of such fully-supervised training settings, a large quantity of expensive and time-consuming pixel-level annotations are required to generate density maps as the supervision. One way to reduce costly labeling is to exploit self-structural information and inner-relations among unlabeled images. Unlike the previous methods utilizing these relations and structural information from the original image level, we explore such self-relations from the latent feature spaces because it can extract more abundant relations and structural information. Specifically, we propose S$^2$FPR which can extract structural information and learn partial orders of coarse-to-fine pyramid features in the latent space for better crowd counting with massive unlabeled images. In addition, we collect a new unlabeled crowd counting dataset (FUDAN-UCC) with 4,000 images in total for training. One by-product is that our proposed S$^2$FPR method can leverage numerous partial orders in the latent space among unlabeled images to strengthen the model representation capability and reduce the estimation errors for the crowd counting task. Extensive experiments on four benchmark datasets, i.e. the UCF-QNRF, the ShanghaiTech PartA and PartB, and the UCF-CC-50, show the effectiveness of our method compared with previous semi-supervised methods. The source code and dataset are available at https://github.com/bridgeqiqi/S2FPR.