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 channel ordering for inputs of multiple Image recognition models? #400

Closed
ermolenkodev opened this issue Jun 2, 2022 · 0 comments
Closed
Assignees
Labels
bug Something isn't working

Comments

@ermolenkodev
Copy link
Collaborator

ImageRecognitionModel.predictObject is a common method for all classification models, and it contains a hardcoded conversion of an input image to BGR format. Meanwhile, most of the models in ModelHub requires input in image in RGB format which is a cause of a drop of classification accuracy.

We should further investigate this issue. If it's indeed an error, we should fix the input format for models that require RGB input, and also input format should be explicitly documented for each model in ModelHub.

@ermolenkodev ermolenkodev added the bug Something isn't working label Jun 2, 2022
@ermolenkodev ermolenkodev self-assigned this Jun 2, 2022
ermolenkodev added a commit to ermolenkodev/KotlinDL that referenced this issue Jun 8, 2022
ermolenkodev added a commit to ermolenkodev/KotlinDL that referenced this issue Jun 23, 2022
ermolenkodev added a commit to ermolenkodev/KotlinDL that referenced this issue Jun 23, 2022
ermolenkodev added a commit to ermolenkodev/KotlinDL that referenced this issue Jul 12, 2022
ermolenkodev added a commit to ermolenkodev/KotlinDL that referenced this issue Jul 12, 2022
ermolenkodev added a commit that referenced this issue Jul 12, 2022
* Fix channels ordering for classification models (#400)

* Add comment for an inputColorMode field (#400)
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
bug Something isn't working
Projects
None yet
Development

No branches or pull requests

1 participant