search

Lornatang / YOLOv3-PyTorch

Pytorch implements yolov3.Good performance, easy to use, fast speed.
Apache License 2.0
32 stars 15 forks source link

when using mobilenet v2, the mAP is rather small and i didn't get the expected result. #6

Closed sddai closed 11 months ago

sddai commented 3 years ago

Instructions To Reproduce the Issue:

  1. what changes you made (git diff) or what code you wrote 
    i only changed the cfg file and weight file to mobilev2 in test.py
  2. what exact command you run: python test.py
  3. what you observed (including full logs): the mAP is rather small and i didn't get the expected result. 
    
           Class    Images   Targets         P         R   mAP@0.5        F1: 100%|████████████████████| 313/313 [00:50<00:00,  6.17it/s]
             all     5e+03  3.58e+04    0.0144    0.0101   0.00985    0.0114
          person     5e+03  1.09e+04    0.0377  0.000551  0.000341   0.00109
         bicycle     5e+03       316     0.443     0.201     0.185     0.277
             car     5e+03  1.67e+03         0         0  3.11e-06         0
       motorbike     5e+03       391         0         0  7.81e-06         0
       aeroplane     5e+03       131         0         0         0         0
             bus     5e+03       261     0.653     0.575       0.6     0.611
           train     5e+03       212         0         0   0.00012         0
           truck     5e+03       352         0         0  9.77e-06         0
            boat     5e+03       475         0         0  1.51e-05         0
   traffic light     5e+03       516         0         0         0         0
    fire hydrant     5e+03        83         0         0  6.15e-05         0
       stop sign     5e+03        84         0         0         0         0
   parking meter     5e+03        59         0         0         0         0
           bench     5e+03       473         0         0  0.000213         0
            bird     5e+03       469         0         0  7.56e-05         0
             cat     5e+03       195   0.00279   0.00513  6.55e-05   0.00362
             dog     5e+03       223    0.0126    0.0224   0.00196    0.0162
           horse     5e+03       305         0         0  3.03e-06         0
           sheep     5e+03       321         0         0         0         0
             cow     5e+03       384         0         0  7.31e-07         0
        elephant     5e+03       284         0         0         0         0
            bear     5e+03        53         0         0         0         0
           zebra     5e+03       277         0         0         0         0
         giraffe     5e+03       170         0         0         0         0
        backpack     5e+03       384         0         0         0         0
        umbrella     5e+03       392         0         0         0         0
         handbag     5e+03       483         0         0         0         0
             tie     5e+03       297         0         0         0         0
        suitcase     5e+03       310         0         0         0         0
         frisbee     5e+03       109         0         0         0         0
            skis     5e+03       282         0         0         0         0
       snowboard     5e+03        92         0         0         0         0
     sports ball     5e+03       236         0         0         0         0
            kite     5e+03       399         0         0         0         0
    baseball bat     5e+03       125         0         0         0         0
  baseball glove     5e+03       139         0         0         0         0
      skateboard     5e+03       218         0         0         0         0
       surfboard     5e+03       266         0         0         0         0
   tennis racket     5e+03       183         0         0         0         0
          bottle     5e+03       966         0         0         0         0
      wine glass     5e+03       366         0         0         0         0
             cup     5e+03       897         0         0         0         0
            fork     5e+03       234         0         0         0         0
           knife     5e+03       291         0         0         0         0
           spoon     5e+03       253         0         0         0         0
            bowl     5e+03       620         0         0         0         0
          banana     5e+03       371         0         0         0         0
           apple     5e+03       158         0         0         0         0
        sandwich     5e+03       160         0         0         0         0
          orange     5e+03       189         0         0         0         0
        broccoli     5e+03       332         0         0         0         0
          carrot     5e+03       346         0         0         0         0
         hot dog     5e+03       164         0         0         0         0
           pizza     5e+03       224         0         0         0         0
           donut     5e+03       237         0         0         0         0
            cake     5e+03       241         0         0         0         0
           chair     5e+03  1.62e+03         0         0         0         0
            sofa     5e+03       236         0         0         0         0
     pottedplant     5e+03       431         0         0         0         0
             bed     5e+03       195         0         0         0         0
     diningtable     5e+03       634         0         0         0         0
          toilet     5e+03       179         0         0         0         0
       tvmonitor     5e+03       257         0         0         0         0
          laptop     5e+03       237         0         0         0         0
           mouse     5e+03        95         0         0         0         0
          remote     5e+03       241         0         0         0         0
        keyboard     5e+03       117         0         0         0         0
      cell phone     5e+03       291         0         0         0         0
       microwave     5e+03        88         0         0         0         0
            oven     5e+03       142         0         0         0         0
         toaster     5e+03        11         0         0         0         0
            sink     5e+03       211         0         0         0         0
    refrigerator     5e+03       107         0         0         0         0
            book     5e+03  1.08e+03         0         0         0         0
           clock     5e+03       292         0         0         0         0
            vase     5e+03       353         0         0         0         0
        scissors     5e+03        56         0         0         0         0
      teddy bear     5e+03       245         0         0         0         0
      hair drier     5e+03        11         0         0         0         0
      toothbrush     5e+03        77         0         0         0         0
/home/dsd/local/venv/lib/python3.8/site-packages/torch/cuda/memory.py:344: FutureWarning: torch.cuda.memory_cached has been renamed to torch.cuda.memory_reserved
warnings.warn(
Inference menory: 1.4 GB.
Speed:
Image size: (416x416) at batch_size: 8
    - Inference 3.7ms.
    - NMS       3.8ms.
    - Total     7.5ms.
Lornatang commented 11 months ago

The latest method fixes the problem of heavy training.