Gait attracts growing interest from researchers due to its advantages as a non-invasive and non-cooperative biometric feature. Current gait-based attribute recognition methods primarily focus on estimating attributes such as gender, age, and emotions. However, there is insufficient attention to diverse gait attributes in various covariate scenarios. In this paper, we design and collect a Richly Annotated benchmark for 15 gait attributes, named RA-GAR, comprising data from 533 individuals with over 120,000 sequences. To our knowledge, RA-GAR represents the largest and most diverse benchmark of gait attributes currently available. Furthermore, to fully leverage the semantic information and enhance attribute-specific local perception, we propose a two-stage CLIP-based method for Gait Attribute Recognition, named CLIP-GAR. Experiments on the RA-GAR and MA-Gait datasets demonstrate the effectiveness of CLIP-GAR, showing significant improvements in mean accuracy and F1 score.

In the field of gait recognition, regarding gait as a set has emerged as a seminal approach, notably eliminating the dependence on template-based input. Although set-based methods offer notable advantages, such as insensitivity to frame order permutations and robustness to varying frame counts, their performance has consistently lagged behind that of sequence-based methods in subsequent studies. In this work, we advocate for treating gait as an unordered set and argue thatthe lack of set context aggregation in frame-level feature extraction is the primary limitation hindering the full potential of set-based gait recognition. To substantiate this claim, we develop a gait-oriented self-attention module and introduce a Gating Mechanism that facilitates set context awareness for each silhouette whilepreserving the permutation-invariant property. Specifically, the context aggregation operates on diverse bins of feature maps, interleaving fine-grained shape and motion details in an almost parameter-free manner. The Gating Mechanism is employed to ensure that frame-level features are not overwhelmed by the aggregated context. Furthermore, the sampling strategy is carefully enhanced to better support set context modeling. Our research demonstrates that set-based gait recognition can achieve state-of-the-art accuracy on in-the-wild benchmarks (77.6% on Gait3D and 81.1% on GREW) while retaining its inherent advantages.

In this work, we present AerialGait, a comprehensive dataset for aerial-ground gait recognition. This dataset comprises 82,454 sequences totaling over 10 million frames from 533 subjects, captured from both aerial and ground perspectives. To align with real-life scenarios of aerial and ground surveillance, we utilize a drone and a ground surveillance camera for data acquisition. The drone is operated at various speeds, directions, and altitudes. Meanwhile, we conduct data collection across five diverse surveillance sites to ensure a comprehensive simulation of real-world settings. AerialGait has several unique features: 1) The gait sequences exhibit significant variations in views, resolutions, and illumination across five distinct scenes. 2) It incorporates challenges of motion blur and frame discontinuity due to drone mobility. 3) The dataset reflects the domain gap caused by the view disparity between aerial and ground views, presenting a realistic challenge for drone-based gait recognition. Moreover, we perform a comprehensive analysis of existing gait recognition methods on AerialGait dataset and propose the Aerial-Ground Gait Network (AGG-Net). AGG-Net effectively learns discriminative features from aerial views by uncertainty learning and clusters features across aerial and ground views through prototype learning. Our model achieves state-of-the-art performance on both AerialGait and DroneGait datasets.

Gait recognition aims to obtain people's identity through body shape and walking posture. Existing gait recognition studies focus on low vertical view recognition, in which the person and the camera are nearly at the same height. Differently, in this work, we focus on gait recognition at high vertical views. To facilitate the research, we propose a new dataset named DroneGait, where the drones are used to collect the gait data. This dataset contains 22 k sequences of 96 subjects taken at different vertical views, varying from about 0$^{\circ }$to 80$^{\circ }$. Furthermore, we evaluate the effectiveness of several state-of-the-art appearance-based and skeleton-based models using our dataset and establish comprehensive baselines. Our results demonstrate that the dataset is challenging and presents significant opportunities to improve existing gait recognition methods. Moreover, we propose a new method called Vertical Distillation, which is based on the feature distillation across different vertical views. Our proposed method substantially outperforms the state-of-the-art models on DroneGait at high vertical views. Cross-vertical-view and cross-domain experiments are also made to explain the importance of gait recognition at high vertical views. Furthermore, we analyze the differences between gait recognition at different vertical views using heatmap visualization techniques. We will make our dataset and code publicly available upon acceptance.

Scanpath refelects the shift of visual attention, therefore prediction of scanpath plays an important role in image analysis and understanding. However traditional scanpath prediction methods ignore the individuality of subjects such as oculomotor bias and other relevant factors. Hence, to make the scanpath prediction more accurate, we incorporate individual traits into a universal scanpath prediction framework for the subject based on the saccade distribution and factor weighting. Experiments demonstrate that our model improves the predicting performance, which proves that individuality is an important factor in scanpath prediction.

Eye movement reflects the shift of overt visual attention. Eye movement trajectories from a group of observers can be expressed by a representative scanpath. The representative scan-path can work as a baseline for studies on scanpath prediction as well as provide useful knowledge about group behavior in psychological studies. In this paper, we propose a new framework to summarize a representative scanpath from individual scanpaths, taking into account the spatial distribution of scan-paths rather than simply treating them as strings of characters. It consists of three steps: extract areas of interest (AOI), remove outliers and summarize scanpaths. In the last step, we develop an algorithm termed Candidate-constrained DTW Barycenter Algorithm (CDBA) by imposing 3 constraints: (1) the components of the representative scanpath must be chosen from candidates (extracted AOIs); (2) the occurrence count of each AOI in the representative scanpath cannot exceed its maximum occurrence count in individual scanpaths; (3) any two contiguous AOIs in the representative scanpath must be contiguous in at least one individual scanpath. The experiments demonstrate that the proposed method outperforms other state-of-the-art scanpath mining methods.