damo-cv/TransReID
A transformer-based deep learning model for object re-identification achieving state-of-the-art on person and vehicle re-ID benchmarks.
Velocity · 7d
+0.5
★ / day
Trend
→steady
star history
TransReID is an official implementation of an ICCV 2021 paper that applies Vision Transformer architecture to object re-identification tasks. The model handles person re-ID and vehicle re-ID using transformer encoders with specialized loss functions and training pipelines. It provides code for training and evaluation on standard benchmarks including Market-1501, MSMT17, DukeMTMC-reID, VehicleID, and VeRi-776.