ValueError: operands could not be broadcast together with shapes (65536,5) (256,256,1)

[GusRoth]

please help me, I dont know how to solve this problem

[filippocastelli]

@GusRoth Fixing the typo reported in #1 solves the problem bro!

[GusRoth]

@GusRoth Fixing the typo reported in #1 solves the problem bro!

thank you！But the problem is still there....
maybe this is because I don't solves the problem?

predictions_smooth = predict_img_with_smooth_windowing(
input_img,
window_size=window_size,
subdivisions=2, # Minimal amount of overlap for windowing. Must be an even number.
nb_classes=nb_classes,
pred_func=(
lambda img_batch_subdiv: model.predict(img_batch_subdiv)
)
)

I delete the img_to_predict_input directly ,because I don't find the function in the author's code ...
Could you tell me how do you solve this function problem? Or it is ok to just np.expands_dims(img,axis =0 ) to shape(batch_size,h,w,nb_classes)?

[Mahi-Mai]

Hello @GusRoth . Did you ever figure this out? If so, would you be willing to share your solutions/code?

[fadishaar84]

@Mahi-Mai @GusRoth @filippocastelli I'm facing the same error, the code throws the same mentioned exception and the proposed solution did not fix it. I appreciate if you could share the solution if you manage to solve it. Thank you.

[Mahi-Mai]

@fadishaar84 Here!

predictions_smooth = predict_img_with_smooth_windowing(
    input_img,
    window_size=window_size,
    subdivisions=2,  # Minimal amount of overlap for windowing. Must be an even number.
    nb_classes=nb_classes,
    pred_func=(
        lambda img_batch_subdiv: model.predict(image_to_neural_input(img_batch_subdiv))
    )
)

Try replacing the value of the last argument (pred_func) with:
(lambda img_batch_subdiv: model.predict(img_batch_subdiv))

I don't know what the heck image_to_neural_input() is supposed to be. It's pretty darned cryptic and it doesn't look like you need it.

Now let me know if you find a good way to speed this process up. On large images it's very slow, but the results are lovely! XD

[bnsreenu]

To make this code usable here are the changes you need to make:

1. replace padx to pady in lines 174 and 212where it starts with... for j in range(0, pady_len-window_size+1.......
2. Change pred_func to... pred_func=(lambda img_batch_subdiv: model.predict((img_batch_subdiv))
3. Provide input as (IMG_WIDTH, IMG_HEIGHT, NUM_CHANNELS). For example provide your large image input as (2048, 1792, 3). Do not provide input at (Num_images, IMG_WIDTH, IMG_HEIGHT, NUM_CHANNELS). The code does that internally by dividing large image into patches. Make sure you preprocess your images just the way you've done during training.

[ac43934]

In addition to the changes suggested by @bnsreenu,

for RGB images in _window_2D function: replace
np.expand_dims(np.expand_dims(wind, 3), 3)
with
np.expand_dims(np.expand_dims(wind, 1), 1)

[Akwasi-Richie]

hi @bnsreenu

having an issue and I would be glad if you can help me out. I encounter the error shown below. however, when I print my input image shape is (5000,5000,3)

File "C:\Users\Richmond\AppData\Local\Temp\ipykernel_26400\2710991319.py", line 92, in
lambda img_batch_subdiv: model.predict((img_batch_subdiv))

File "C:\Users\Richmond\AppData\Local\Temp\ipykernel_26400\2283882284.py", line 272, in predict_img_with_smooth_windowing
plt.imshow(prd)

File "C:\Users\Richmond\anaconda3\envs\tensorflow\lib\site-packages\matplotlib_api\deprecation.py", line 456, in wrapper
return func(*args, **kwargs)

File "C:\Users\Richmond\anaconda3\envs\tensorflow\lib\site-packages\matplotlib\pyplot.py", line 2647, in imshow
**kwargs)

File "C:\Users\Richmond\anaconda3\envs\tensorflow\lib\site-packages\matplotlib_api\deprecation.py", line 456, in wrapper
return func(*args, **kwargs)

File "C:\Users\Richmond\anaconda3\envs\tensorflow\lib\site-packages\matplotlib_init_.py", line 1412, in inner
return func(ax, *map(sanitize_sequence, args), **kwargs)

File "C:\Users\Richmond\anaconda3\envs\tensorflow\lib\site-packages\matplotlib\axes_axes.py", line 5488, in imshow
im.set_data(X)

File "C:\Users\Richmond\anaconda3\envs\tensorflow\lib\site-packages\matplotlib\image.py", line 716, in set_data
.format(self._A.shape))

TypeError: Invalid shape (5000, 5000, 2) for image data.

Thank you