Gait dataset with Objective and Subjective Interferences (GOSI)

The GOSI dataset is an indoor gait corpus collected from 160 pedestrians to enable rigorous evaluation of methods under objective and subjective interferences. Data were acquired with four ORBBEC cameras positioned at 0\u00b0, 30\u00b0, 60\u00b0, and 90\u00b0, with camera heights from 1.9 to 2.3 meters to emulate typical surveillance viewpoints. The collection introduces objective variability from lighting, clothing, and camera height, together with subjective behaviors such as holding a hat or keeping the arms crossed. Walking conditions are organized into normal walking, walking with a hat, and an overlong coat with arms crossed. For every participant two sequences are recorded for each condition across all views, yielding 24 sequences per individual and supporting systematic analysis of cross view robustness and interference tolerance.