Large-Scale Spatio-Temporal Person Re-identification: Algorithm and Benchmark

LaST is a large-scale benchmark consisting of 10,862 identities and more than 228k images. The activity scope of pedestrians in LaST is quite large, in different blocks of a city, or from one city to another, or even, some people cross different countries. The time ranges from several days to several months, therefore, clothes changing is common for pedestrians. The images vary from day to night, from spring to winter.

Data Collection

Statistical Information

Clothing labels

Sample Images

Cloth-changing Samples

Agreement

The benchmark is available for non-commercial research purposes only.