Closed rehoyt closed 2 years ago
Model performs just the modelling part (i.e. logistic regression). It does not embed the images. You are missing Image Embedding in between Import Images and Predictions.
Thanks. I added the image embedding widget to the workflow but the predictions are extremely low, whereas they were high in the model I saved and loaded for the new images. I used the Google Inception 3 CNN. See below
[image: image.png]
On Mon, Jan 10, 2022 at 2:51 AM Ajda @.***> wrote:
Model performs just the modelling part (i.e. logistic regression). It does not embed the images. You are missing Image Embedding in between Import Images and Predictions.
— Reply to this email directly, view it on GitHub https://github.com/biolab/orange3/issues/5772#issuecomment-1008656485, or unsubscribe https://github.com/notifications/unsubscribe-auth/AGDWVUTZXWNBLM6SLWE7N7LUVKMZZANCNFSM5LSMQVJA . Triage notifications on the go with GitHub Mobile for iOS https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675 or Android https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub.
You are receiving this because you authored the thread.Message ID: @.***>
--
Robert (Bob) Hoyt MD, FACP, FAMIA, ABPM-CI
Associate Clinical Professor, Department of Internal Medicine
Virginia Commonwealth University
Richmond, VA
CAPT (Ret) USN
InformaticsEducation.org http://InformaticsEducation.org @.***
Cell: 850-384-5235
I chose another image dataset and everything was fine
I understand how the save model and load model widgets work with numerical data, but I question whether they work with image data. I uploaded ECG images (normal vs atrial fibrillation)(separate subfolders) to the image embedding widget and created a model with the logistic regression widget connected to the image embedding widget and the test and score widget. I saved the model as a pickle file.
I then uploaded a new folder of normal vs atrial fibrillation images to an import images widget and connected that to a prediction widget and the latter to the load model widget. The problem is that I don't believe the results or am I doing something wrong? The predictions vary between 0.46 and 0.56 for atrial fibrillation and for a normal ECG. See workflow below