Is your feature request related to a problem?
Yes. Users can perform general image processing, but they cannot easily identify smiling regions or build facial-expression-style classical computer vision demos using the existing operator set. This limits the usefulness of the Detection area for facial feature analysis and makes it harder to create simple face-and-smile pipelines in the app.
Describe the solution you'd like
Add a Smile Detection operator under a Detection category, backed by OpenCV Haar cascades on the backend. The implementation should detect smiles in a practical way by first detecting faces, then searching only within the lower face region for likely smiles, and drawing smile bounding boxes on the output image.
Describe alternatives you've considered
Running the smile Haar cascade across the full image directly, but that tends to produce many false positives and performs worse than limiting detection to the lower region of detected faces.
Using a more modern detector such as MediaPipe, DNN-based facial landmark models, or expression recognition models, which would likely improve accuracy, but would add more complexity and dependencies than a classical OpenCV-based first implementation.
Is your feature request related to a problem?
Yes. Users can perform general image processing, but they cannot easily identify smiling regions or build facial-expression-style classical computer vision demos using the existing operator set. This limits the usefulness of the Detection area for facial feature analysis and makes it harder to create simple face-and-smile pipelines in the app.
Describe the solution you'd like
Add a Smile Detection operator under a Detection category, backed by OpenCV Haar cascades on the backend. The implementation should detect smiles in a practical way by first detecting faces, then searching only within the lower face region for likely smiles, and drawing smile bounding boxes on the output image.
Describe alternatives you've considered
Running the smile Haar cascade across the full image directly, but that tends to produce many false positives and performs worse than limiting detection to the lower region of detected faces.
Using a more modern detector such as MediaPipe, DNN-based facial landmark models, or expression recognition models, which would likely improve accuracy, but would add more complexity and dependencies than a classical OpenCV-based first implementation.