文章

Day30 - Baseline在FLIR上的表现

以yolov5n/s/m和yolov8n/s作为baseline,在FLIR上训练200epochs

  • Yolov5n

    Parameters: 2.655M

    ClassPrecisionRecallmAP50mAP50-95
    all0.6040.4450.4870.293
    person0.8320.6050.740.415
    bike0.5170.4350.4110.247
    car0.8550.7420.8350.588
    motor0.7570.5660.6340.339
    bus0.7710.4750.6110.425
    truck0.2410.3260.1520.102
    light0.7640.3730.4840.22
    other vehicle0.09470.04160.0310.01
  • Yolov5s

    Parameters: 9.153M

    ClassPrecisionRecallmAP50mAP50-95
    all0.6370.4950.5530.34
    person0.8660.650.7880.467
    bike0.6050.5180.4970.312
    car0.880.7770.8670.628
    motor0.7930.6180.7550.379
    bus0.7720.4530.6390.461
    truck0.3070.4350.270.183
    light0.7970.4590.5730.272
    other vehicle0.07980.04760.03690.0169
  • Yolov5m

    Parameters: 25.111M

    ClassPrecisionRecallmAP50mAP50-95
    all0.6760.540.5790.371
    person0.8710.6770.8080.498
    bike0.5970.5650.550.359
    car0.8870.7980.8810.651
    motor0.7710.6550.7310.418
    bus0.8120.5530.6910.509
    truck0.4070.4570.2830.191
    light0.8040.5030.6070.303
    other vehicle0.2610.1110.08160.0423
  • Yolov8n

    Parameters: 3.157M

    ClassPrecisionRecallmAP50mAP50-95
    all0.60.4330.480.288
    person0.8350.6140.7430.418
    bike0.4980.4120.4040.246
    car0.8590.7470.8380.595
    motor0.7960.5690.610.314
    bus0.780.4550.6050.413
    truck0.1950.2390.1340.0827
    light0.7230.3810.4730.221
    other vehicle0.1110.04760.03670.0171
  • Yolov8s

    Parameters: 11.167M

    ClassPrecisionRecallmAP50mAP50-95
    all0.6260.5220.550.347
    person0.8340.6810.790.474
    bike0.530.5590.510.313
    car0.8410.8080.870.633
    motor0.820.6180.700.399
    bus0.7680.5350.660.477
    truck0.2780.4130.230.162
    light0.7650.4680.570.272
    other vehicle0.1730.09520.0870.0469
本文由作者按照 CC BY 4.0 进行授权