Demonstrating an Web based API which will label images with existing model and can be retrained to enhance the prediction accuracy in UI itself. The portal uses Active Learning & transfer learning techniques Please check out the app at this link- https://ankan-mazumdar-active-learning-stream-app-kbbcqv.streamlit.app/
NOTE- the model is currently trained inly on CIFAR 10 image dataset, please upload any CIFAR 10 image only. Please follow the below flowchart to get desired predictions-
Here, as HORSE is incorrectly predicted as CAT, user is manually submitting the label as HORSE from dropdown as then retraining it
The model can be retrained further with more number of iterations to increase the prediction accuracy.