Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Wrong output for custom images #7

Open
liuhuihuii opened this issue Dec 30, 2018 · 6 comments
Open

Wrong output for custom images #7

liuhuihuii opened this issue Dec 30, 2018 · 6 comments

Comments

@liuhuihuii
Copy link

I solved the problem with the image size last night and I should thank you for that, however, after I trained the network for 24 epochs, the prediction of custom images are still far from correct. The percentage of each label it gives for custom images is around 25%, while testing random images from the data set gives very definite answer (50% for one label). Is it because of overfitting or not enough epochs? Is there any advice to improve the correctness for custom images?

@rishiswethan
Copy link
Owner

rishiswethan commented Dec 30, 2018 via email

@liuhuihuii
Copy link
Author

But when I use your pretrianed mode 'my_model3.h5' t to predict my own cunstom image named 'InvasiveC2_2048x1536.jpg' ,then output is like:
Average from all crops

Benign : 20.6134%
InSitu : 31.7345%
Invasive : 20.0719%
Normal : 27.5802%

@rishiswethan
Copy link
Owner

rishiswethan commented Dec 30, 2018 via email

@liuhuihuii
Copy link
Author

I downloaded them from the internet. Maybe they are not labeled correctly, but I trust wiki anyways

@liuhuihuii
Copy link
Author

For example, this image is from Wikipedia and it says normal on the site
the output is
Benign : 29.1999%
InSitu : 21.6953%
Invasive : 26.8959%
Normal : 22.2089%
all percentage is close to 25%
normalc1_2048x1536

@rishiswethan
Copy link
Owner

rishiswethan commented Dec 30, 2018 via email

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

2 participants