Under the Track3 folder, there are 3 components:
- 1_Multi-Camera Vehicle Tracking and Re-identification: Multi-camera vehicle tracking based on a fusion of histogram-based adaptive appearance models, DCNN features, detected car types and traveling time information
- 2_YOLO_LP: Detection of license plate from each cropped vehicle image based on YOLOv2 with our trained model provided We strongly encourage users to try the latest YOLOv4 object detector instead.
- 3_LP_COMP: Comparison of license plates under low resolution
Detailed description of each package is given in each subfolder.
The output of 1_Multi-Camera Vehicle Tracking and Re-identification is the similarity scores between pairs of vehicles for comparison. It can be converted into a distance score by inverse proportion. The output of 3_LP_COMP is the distance score between each two license plates. The final distance score between two vehicles is the multiplication of the above two distance scores.