在python中显示原始图像像素而不是蒙版

Question

I have a deep learning model which returns to me an array which when plotted like this

res = deeplab_model.predict(np.expand_dims(resized2,0))
labels = np.argmax(res.squeeze(),-1) #remove single dimension values, gives the indices of maximum values in the array  
plt.imshow(labels[:-pad_x])

(the last line above just removes some unclear lines before plotting them)

looks like this

original image is like this

when I do the

print(labels[labels>0])
print(labels.shape)
print(len(labels))

I get this

[12 12 12 ... 12 12 12]
(512, 512)
512

I want to show the colored pixels in the original image where mask appears and turn everything else to black (or blur or some other color I'll choose), how can I do that?

Answer 1

It's not entirely clear how the labels array works here. Assuming that it contains values greater than zero where the cat and dog are, you can create the masked image with something like,

mask = lables > 0
newimage = np.zeros(image.shape)
newimage[mask] = image[mask]

where I've create a zero image based on the original and set the original pixels where the labels are greater than zero.

Answer 2

我能够扭转这一局面并实现我想要的目标

mask = labels[:-pad_x] == 0 resizedOrig = cv2.resize(frame, (512,384)) resizedOrig[mask] = 0