2019-CVPR-AIC-Track-1-UWIPL

1. Single Camera Tracking

Please download the source code and follow the instruction from https://github.com/GaoangW/TNT/tree/master/AIC19.

2. Deep Feature Re_identification

Please download the source code and follow the instruction from https://github.com/ipl-uw/2019-CVPR-AIC-Track-2-UWIPL. Create video2img folder in the downloaded project (i.e., Video-Person-ReID/video2img/). Put and run python crop_img.py in the same folder in the downloaded dataset (i.e., aic19-track1-mtmc/test). You need to creat a folder track1_test_img in the same path (i.e., aic19-track1-mtmc/test/track1_test_img). After that, create a folder track1_sct_img_test_big and run python crop_img_big.py. Then, create a folder log in the dowanloaded project (i.e., Video-Person-ReID/log) and put the downloaded model file of track1 ReID in this folder. Finally, run python Graph_ModelDataGen.py to obtain the feature files (q_camids3_no_nms_big0510.npy, qf3_no_nms_big0510.npy and q_pids3_no_nms_big0510.npy).
The code is based on Jiyang Gao's Video-Person-ReID [code].

3. Trajectory-Based Camera Link Models

Put the feature files (q_camids3_no_nms_big0510.npy, qf3_no_nms_big0510.npy and q_pids3_no_nms_big0510.npy) in the Transition-Model folder of this project. Then, run python main_in_transition_matrix.py. Then, find the results in Transition-Model/transition_data /ICT-no_nms_big510/. The filename of the result should look like ict_greedy_iter1600_10.559_654_498.txt (1600 means the number of iteration, 10.559 represents the distance of embeddings at current iteration, 654 means the maximum globelID assigned to cross-camera vehicles, and 498 is the number of unique globelID of cross-camera vehicles after merging).

4. NMS

Put the output result (e.g. ict_greedy_iter1600_10.559_654_498.txt) from previous step in NMS folder. Then, run python NMS_filter.py to get the final result (e.g. ict_greedy_iter1600_10.559_654_498_big.txt) of Track1.