Automatic Learning of Gait Signatures for People Identification

Abstract

This work targets people identification in video based on the way they walk (i.e. gait). While classical methods typically derive gait signatures from sequences of binary silhouettes, in this work we explore the use of convolutional neural networks (CNN) for learning high-level descriptors from low-level motion features (i.e. optical flow components). We carry out a thorough experimental evaluation of the proposed CNN architecture on the challenging TUM-GAID dataset. The experimental results indicate that using spatio-temporal cuboids of optical flow as input data for CNN allows to obtain state-of-the-art results on the gait task with an image resolution eight times lower than the previously reported results (i.e. 80 × 60 pixels).

Deep Learning-Based Gait Recognition Using Smartphones in the Wild

“…CNN can also be combined with traditional machine learning methods such as PCA, MCA [63], Bayesian classifier [64] and SVM [65]. In [65], CNN was used as a feature extractor, then the extracted features were classified by SVM. At the same time, there are also some researches using CNN to extract three-dimensional data consisting of images and optical flow information for gait and activity recognition [65], [66].…”

Reinforced Temporal Attention and Split-Rate Transfer for Depth-Based Person Re-identification

“…The small sample size problem is especially acute in person re-identification from temporal sequences [26,86,9,54,110], as the feature dimensionality increases linearly in the number of frames that are accumulated compared to the singleshot representations. On the other hand, explicitly modeling temporal dynamics and using multiple frames help algorithms to deal with noisy measurements, occlusions, adverse poses and lighting.…”

Gait Recognition Based on Convolutional Neural Networks

“…We started from the architecture used in (Castro et al, 2016) that is based on the CNN-M architecture from (Chatfield et al, 2014) and used model based on this architecture as a baseline. The only modification we made is using Batch Normalization layer (Ioffe and Szegedy, 2015) instead of Local Response Normalization used initially.…”

“…Unlike these methods, we extract gait features from motion itself training convolutional neural network (CNN). CNN models are very successful in many computer vision problems, but their first application to gait recognition was made not long ago in (Castro et al, 2016). And we are going to push the performance further.…”