Testing problem

I am facing the following error when I eval the pre-trained model on CityPersons dataset.

  File "./tools/test_city_person.py", line 230, in <module>
    main()
  File "./tools/test_city_person.py", line 222, in main
    MRs = validate('datasets/CityPersons/val_gt.json', args.out)
  File "/home/mingzhi/Downloads/Pedestron/tools/cityPerson/eval_demo.py", line 13, in validate
    cocoEval = COCOeval(cocoGt, cocoDt, 'bbox')
  File "/home/mingzhi/Downloads/Pedestron/tools/cityPerson/eval_MR_multisetup.py", line 76, in __init__
    self.params = Params(iouType=iouType) # parameters
  File "/home/mingzhi/Downloads/Pedestron/tools/cityPerson/eval_MR_multisetup.py", line 525, in __init__
    self.setDetParams()
  File "/home/mingzhi/Downloads/Pedestron/tools/cityPerson/eval_MR_multisetup.py", line 501, in setDetParams
    self.recThrs = np.linspace(.0, 1.00, np.round((1.00 - .0) / .01) + 1, endpoint=True)
  File "<__array_function__ internals>", line 6, in linspace
  File "/home/mingzhi/anaconda3/envs/pedestron/lib/python3.7/site-packages/numpy/core/function_base.py", line 113, in linspace
    num = operator.index(num)
TypeError: 'numpy.float64' object cannot be interpreted as an integer

When I train the model and eval my own trained models, I am having this problem: File "./tools/test_city_person.py", line 230, in <module> main() File "./tools/test_city_person.py", line 222, in main MRs = validate('datasets/CityPersons/val_gt.json', args.out) File "/home/mingzhi/Downloads/Pedestron/tools/cityPerson/eval_demo.py", line 11, in validate cocoDt = cocoGt.loadRes(dt_path) File "/home/mingzhi/Downloads/Pedestron/tools/cityPerson/coco.py", line 313, in loadRes if 'caption' in anns[0]: IndexError: list index out of range

How can I solve the above two problems?? I know for the second one, it seems the model does not detect, but why?

[msha096]

Numpy problem solution update: Just in case someone has the same problem with NumPy version. You could either downgrade NumPy to lower version (lower than 1.18) (I failed with this solution) To use with NumPy >=1.18, you could try pip install -U 'git+https://github.com/cocodataset/cocoapi.git#subdirectory=PythonAPI'(this solution does not work for me but it works for someone using MMDetection)

The solution works for me is: LOC 501 of the file /Pedestron/tools/cityPerson/eval_MR_multisetup.py, change self.recThrs = np.linspace(.0, 1.00, np.round((1.00 - .0) / .01) + 1, endpoint=True) into self.recThrs = np.linspace(.0, 1.00, 101, endpoint=True)

Cheers

[123dddd]

Try 20-40 epochs.

The LR is too large for one GPU. And that is the reason. After reducing LR, I got around 22% MR on reasonable when I train 12 epochs. How many epochs did you train to get around 14% MR ?

Hi, could you pls show the config here? I also have to choose a small LR for the training, the default LR is large for one GPU case.