Skip to content
Advertisement

I want to train the CSRNet model on UCF_CC_50 dataset which has total number of images 50. and face a problem like

I want to train the CSRNet model on UCF_CC_50 dataset but occurring this problem

KeyError  Traceback (most recent call last)    <ipython-input-11-78e054690de5> in <module>  
     4     img= plt.imread(img_path)
     5     k = np.zeros((img.shape[0],img.shape[1]))
***----> 6     gt = mat["image_info"][0,0][0,0][0]***
     7     for i in range(0,len(gt)):
     8         if int(gt[i][1])<img.shape[0] and int(gt[i][0])<img.shape[1]:

**KeyError: 'image_info'**
----------

enter for img_path in img_paths:
    print (img_path)
    mat = io.loadmat(img_path.replace('.jpg','.mat').replace('images','ground_truth').replace('IMG_','GT_IMG_'))
    img= plt.imread(img_path)
    k = np.zeros((img.shape[0],img.shape[1]))
    gt = mat["image_info"][0,0][0,0][0]
    for i in range(0,len(gt)):
        if int(gt[i][1])<img.shape[0] and int(gt[i][0])<img.shape[1]:
            k[int(gt[i][1]),int(gt[i][0])]=1
    k = gaussian_filter_density(k)
    with h5py.File(img_path.replace('.jpg','.h5').replace('images','groundtruth'), 'w') as hf:
            hf['density'] = kcode here

---------

The file path is

C:UsersGigabyte pcDesktopCOUNTINGCSRNet-pytorch-masterUCF_CC_50part_A_final/train_dataimagesIMG_1.jpg

Advertisement

Answer

Your code does not comply with the structure of the annotation file you are trying to read. Annotations in UCF-50 CC dataset can simply be read by getting the values of the key “annPoints”.

You could apply the following changes to your code to read the x and y coordinates of pointwise human-head annotations:

4 img= plt.imread(img_path)
5 k = np.zeros((img.shape[0],img.shape[1]))
6 gt = mat["annPoints"]
7 for i in range(0,len(gt)):
User contributions licensed under: CC BY-SA
8 People found this is helpful
Advertisement