The goal with this release is to automatize the whole process of fetching and loading ReID models. Multi-cam tracking possibility was added by @hdnh2006 in https://github.com/mikel-brostrom/Yolov5_StrongSORT_OSNet/pull/284

Important updates

Multiple object-tracking on multiple simultaneous streams Tracking with the --save-vid flag on folders containing images now generate an .mp4 Automatic download of ReID models trained on different datasets Easier experimentation by setting the result folder name to the yolo and deep sort models used. The following can now be saved under this folder:

• Tracking with --save-crop flag saves crops associated to each class and ID for each stream
• Tracking with the --save-vid generates an .mp4 for each stream
• Tracking with the --save-txt saves a txt files for each stream

Bug fixes

Loading of custom ReID model by specifying the path to it

MOT16 Train evaluation results

Relevant changed/used hparams: imgz 1280, crowdhuman_yolov5m, OSNet_x0_25_msmt17, alpha=0. Notice that none of the models used during the evaluation has ever seen any of the MOT16 data.

HOTA: rep_1280-pedestrian          HOTA      DetA      AssA      DetRe     DetPr     AssRe     AssPr     LocA      RHOTA     HOTA(0)   LocA(0)   HOTALocA(0)
MOT16-02                           33.472    36.215    31.143    38.083    76.454    33.609    74.943    81.576    34.395    42.156    77.46     32.654    
MOT16-04                           58.508    58.422    59.167    63.228    76.489    63.097    79.334    81.143    61.096    77.915    75.329    58.693    
MOT16-05                           39.479    36.937    42.304    39.621    73.848    49.756    70.33     81.457    40.932    51.081    76.254    38.952    
MOT16-09                           51.897    56.914    47.372    65.839    74.434    51.6      79.236    85.818    55.841    64.133    82.232    52.737    
MOT16-10                           46.407    49.419    43.764    53.425    74.242    48.223    74.634    80.347    48.344    62.128    75.134    46.679    
MOT16-11                           58.33     59.366    57.502    70.96     73.111    65.957    77.009    86.464    63.869    69.923    82.949    58        
MOT16-13                           44.229    41.439    47.558    44.837    72.748    51.187    77.022    80.409    46.127    57.94     75.221    43.583    
COMBINED                           51.286    50.839    52.219    55.475    75.31     56.777    78.178    81.874    53.742    66.59     76.683    51.063    

CLEAR: rep_1280-pedestrian         MOTA      MOTP      MODA      CLR_Re    CLR_Pr    MTR       PTR       MLR       sMOTA     CLR_TP    CLR_FN    CLR_FP    IDSW      MT        PT        ML        Frag      
MOT16-02                           43.773    78.642    44.44     47.126    94.608    16.667    51.852    31.481    33.707    8404      9429      479       119       9         28        17        366       
MOT16-04                           69.634    78.654    69.803    76.233    92.221    48.193    36.145    15.663    53.362    36254     11303     3058      80        40        30        13        536       
MOT16-05                           42.754    78.816    43.532    48.592    90.569    17.6      58.4      24        32.461    3313      3505      345       53        22        73        30        229       
MOT16-09                           61.119    84.468    61.746    75.1      84.903    52        48        0         49.454    3948      1309      702       33        13        12        0         107       
MOT16-10                           60.643    77.175    61.244    66.602    92.554    31.481    55.556    12.963    45.441    8204      4114      660       74        17        30        7         580       
MOT16-11                           63.571    84.937    63.898    80.477    82.918    53.623    36.232    10.145    51.449    7383      1791      1521      30        37        25        7         148       
MOT16-13                           51.022    76.912    51.572    56.603    91.838    28.037    42.056    29.907    37.954    6481      4969      576       63        30        45        32        277       
COMBINED                           59.955    79.281    60.364    67.013    90.974    32.495    47.002    20.503    46.07     73987     36420     7341      452       168       243       106       2243      

Identity: rep_1280-pedestrian      IDF1      IDR       IDP       IDTP      IDFN      IDFP      
MOT16-02                           43.742    32.765    65.777    5843      11990     3040      
MOT16-04                           73.02     66.69     80.678    31716     15841     7596      
MOT16-05                           55.116    42.344    78.923    2887      3931      771       
MOT16-09                           62.764    59.14     66.86     3109      2148      1541      
MOT16-10                           61.175    52.598    73.093    6479      5839      2385      
MOT16-11                           66.777    65.795    67.79     6036      3138      2868      
MOT16-13                           60.323    48.751    79.099    5582      5868      1475      
COMBINED                           64.31     55.841    75.807    61652     48755     19676     

Count: rep_1280-pedestrian         Dets      GT_Dets   IDs       GT_IDs    
MOT16-02                           8883      17833     142       54        
MOT16-04                           39312     47557     138       83        
MOT16-05                           3658      6818      134       125       
MOT16-09                           4650      5257      57        25        
MOT16-10                           8864      12318     118       54        
MOT16-11                           8904      9174      157       69        
MOT16-13                           7057      11450     131       107       
COMBINED                           81328     110407    877       517

Performance boost coming from evaluating on 1280 image size