Object arrays cannot be loaded when allow_pickle=False

[andrazvonar]

I'm having a issue using the test_on_images Jupyter Notebook. I'm running into an issues with the np.load due to in not expecting a pickle object. Is there a workaround for this?

---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
<ipython-input-4-baf688e75413> in <module>
      1 img = Image.open('images/office1.jpg')
----> 2 bounding_boxes, landmarks = detect_faces(img)
      3 show_bboxes(img, bounding_boxes, landmarks)

~/Documents/šola/diploma/mtcnn-pytorch/src/detector.py in detect_faces(image, min_face_size, thresholds, nms_thresholds)
     23 
     24     # LOAD MODELS
---> 25     pnet = PNet()
     26     rnet = RNet()
     27     onet = ONet()

~/Documents/šola/diploma/mtcnn-pytorch/src/get_nets.py in __init__(self)
     53         self.conv4_2 = nn.Conv2d(32, 4, 1, 1)
     54 
---> 55         weights = np.load('src/weights/pnet.npy')[()]
     56         for n, p in self.named_parameters():
     57             p.data = torch.FloatTensor(weights[n])

~/opt/anaconda3/envs/diploma/lib/python3.8/site-packages/numpy/lib/npyio.py in load(file, mmap_mode, allow_pickle, fix_imports, encoding)
    450                 return format.open_memmap(file, mode=mmap_mode)
    451             else:
--> 452                 return format.read_array(fid, allow_pickle=allow_pickle,
    453                                          pickle_kwargs=pickle_kwargs)
    454         else:

~/opt/anaconda3/envs/diploma/lib/python3.8/site-packages/numpy/lib/format.py in read_array(fp, allow_pickle, pickle_kwargs)
    737         # The array contained Python objects. We need to unpickle the data.
    738         if not allow_pickle:
--> 739             raise ValueError("Object arrays cannot be loaded when "
    740                              "allow_pickle=False")
    741         if pickle_kwargs is None:

ValueError: Object arrays cannot be loaded when allow_pickle=False

[andrazvonar]

I fixed the issue by adding allow_pickle=True to np.load calls in get_nets.py. I can submit a pull request for this.

[DevashishPrasad]

weights = np.load('mtcnn_pytorch/src/weights/pnet.npy',allow_pickle=True)[()]

I did the same but still get the error

[knightlibra]

weights = np.load('mtcnn_pytorch/src/weights/pnet.npy',allow_pickle=True)[()]

I did the same but still get the error Maybe you should try to downgrade your numpy version to conda install numpy=1.16.2 or 1.16.1

[DevashishPrasad]

Thank you for the response knightlibra

I was running the script in Google Colab, I refreshed the runtime and error was gone.

It seems the changes were not reflected in the same session. Closing the session and connecting again solved the issue.

[xuyingzhongguo]

weights = np.load('mtcnn_pytorch/src/weights/pnet.npy',allow_pickle=True)[()]

I did the same but still get the error

There are three places you need to modify.