To facilitate gait recognition research, we collect the Front-View Gait (FVG) database in the course of two years, 2017 and 2018. To better protect the privacy of the subjects, we create a version of the database called FVG-B where the face area is blurred to the extent that state-of-the-art face recognition algorithms fail to recognize the subject. Compared to other view angles in gait recognition, frontal-view walking is a more challenging problem since it contains minimal gait cues. FVG-B includes significant variations, e.g., walking speed, carrying, and clothing from frontal view angles.
FVG-B provides frontal walking videos from 226 subjects. In addition, 12 of them were collected twice from the year 2017 and 2018, in total 2,856 videos. The videos were captured by the camera Logitech C920 Pro Webcam or GoPro Hero 5 on a tripod at the height of 1.5 meters at 1, 080 × 1, 920 resolution with the average length of 10 seconds. More details can be found in our paper section 4.
In blurring the dataset, we apply Gaussian filter to the face of the subjects with the smallest kernel font-size such that the cosine similarity between the blurred face and the gallery face images output by the state-of-the-art AdaFace is below a threshold of 0.2.
FVG-B is collected in three sessions. In session 1, in 2017, videos from 147 subjects(#1 to 147) are collected with four variations (normal walking, slow walking, fast walking, and carrying status). In session 2, in 2018, videos from additional 79 subjects(#148 to 226) are collected. Variations are normal, slow or fast walking speed, clothes or shoes change, and twilight or cluttered backgrounds. Finally, in session 3, we collect repeated 12 subjects(#1,2,4,7,8,12,13,17,31,40,48,77) in the year 2018 for the extreme challenging test with the same setup as section 1. The purpose is to test how time gaps affect gait, along with changes in clothes/shoes or walking speed.
Files and Naming
All the video files are released as PNG frames ordered by frame index, .e.g. a name ending with 0010.png represents frame number 10. The file structure of FVG-B is demonstrated below.
+-- README.txt +-- s // Session, RGB frames | +-- iii // Subject ID | | +-- tt // Sequence number | | | +-- ss_iii_x_y_0001.png | | | +-- ss_iii_x_y_0002.png | | | +-- ... | | +-- ... | +-- ... +-- ss_SIL // Silhouettes extracted using MASK R-CNN | +-- iii // Subject ID | | +-- tt // Sequence number | | | +-- ss_iii_x_y_0001.png | | | +-- ss_iii_x_y_0002.png | | | +-- ... | | +-- ... | +-- ... where - s: Session number, 1 or 2 - iii: Subject ID, 001-147 for session 01 and 001-077 for session 02 - tt: Sequence number, 01-12 - x: Variations. - 1: normal - 2: fast - 3: slow - 4: carrying bag/hat - 5: change clothes - 6: multiple people - y: view angle - 1: -45 degrees - 2: 0 degrees - 3: 45 degrees.
Evaluation protocols can be found here.