When I try to upload the image of a bear and click classify. I get this
error
Your post doesn’t show where/how btn_upload is set, which makes it hard to even guess and the issue. Can you link to the whole notebook?
What do you get from…
type(btn_upload)
and…
dir(btn_upload)
I just encountered the same situation. After some tests, here’s my solution:
import matplotlib.pyplot as plt
from PIL import Image
from io import BytesIO
img = Image.open(BytesIO(btn_upload.value[-1].content))
plt.imshow(img)
The type of btn_upload.value[-1].content is a memoryview and therefore I searched for how to load images from memory and then came up with above code snippet.
Edit: For prediction in the later code block, the image format will cause problem and I had to convert it using img = np.array(img).
1 Like
Share how did I find the solution.
As you mentioned, I used dir(btn_upload) to find value and then the content. Also, type is a great tool to work along with dir. 
1 Like
Now my notebook is running but ,I am getting this error can you help me resolve it. If you need any more information please let me know
I can share my notebook as well
It’s the same problem as indicated by the error message.
Original code:
def on_click_classify(change):
img = PILImage.create(btn_upload.data[-1])
out_pl.clear_output()
with out_pl: display(img.to_thumb(128,128))
pred,pred_idx,probs = learn_inf.predict(img)
lbl_pred.value = f'Prediction: {pred}; Probability: {probs[pred_idx]:.04f}'
Modify it as the follows:
def on_click_classify(change):
img = Image.open(BytesIO(btn_upload.value[-1].content)) # change this
out_pl.clear_output()
with out_pl: display(img.to_thumb(128,128))
img = np.array(img) # and add this
pred,pred_idx,probs = learn_inf.predict(img)
lbl_pred.value = f'Prediction: {pred}; Probability: {probs[pred_idx]:.04f}'
2 Likes
Thanks a lot it worked can you please tell me what caused it, why it was not working properly and how can i deploy it using voila
You could remove that line and check the error message.
I can’t remember the details but it’s something about the image format for the learn_inf.predict() function. The img = np.array(img) convert the image to a numpy array which is acceptable to the learn_inf.predict() function.
Yes. Somehow the kernel still takes old ‘on_click_classify’ function into account and hence the error. Try shutting down the kernel, restart it and then run all the code. The error should disapper.
I encountered the same issue when I run the notebook locally. However, the same notebook works fine with the Google colab. Thus I suppose it is somehow related to the version of ipywidgets.
Checking the doc of ipywidgets, it says
Changes in *ipywidgets 8*:
The `FileUpload` changed significantly in ipywidgets 8:
*The `.value` traitlet is now a list of dictionaries, rather than a dictionary mapping the uploaded name to the content. To retrieve the original form, use `{f["name"]: f.content.tobytes() for f in uploader.value}`.
* The `.data` traitlet has been removed. To retrieve it, use `[f.content.tobytes() for f in uploader.value]`.
Thus, I changed the one signle line, now it works perfert. Hope such info is useful for others
#img = PILImage.create(btn_upload.data[-1])
img = PILImage.create(btn_upload.value[0].content.tobytes())
3 Likes