MSU scientists, in collaboration with Tevian, have developed a new method for recognizing people by silhouette — DynaMix. This was reported by the university's press service.
Re-identification (silhouette-based person recognition) involves searching for similar people across different surveillance cameras at different times.
Re-identification algorithms are used in "smart" cities and intelligent video analytics systems, for example, when searching for criminals by their silhouettes or improving business metrics (analyzing queues and building heatmaps of customer movement).
The new approach is based on training the model by "mixing" simpler images of people into multi-camera data. The development has doubled the quality of re-identification due to stylistic diversity.
Earlier, www1.ru reported that face recognition technology is being implemented to monitor workers at construction sites in Moscow.
