Seeking Feedback for Developing Transfer Learning Property Editor Project

Open Source Development
1

Hi, everyone! I am rino
I applied to GSoC for the project "Property Editors for Transfer Learning ML Models" . Although my proposal didn't get selected, the mentor ewpatton for the project has been incredibly supportive and offered to guide me . So, we're going to build it anyway!

Background / context:

App Inventor has extensions like Personal Image Classifier (PIC) and Personal Audio Classifier (PAC) that use external websites to build and train models. To use these models in App Inventor, you have to export and import them, which is a bit cumbersome.

PIC: https://classifier-appinventor-mit-edu.ezproxy.canberra.edu.au/
PAC: https://mv-ezproxy-com.ezproxy.canberra.edu.au/

Goals:

Here are a list of things we are trying to accomplish

  1. to create property editors for transfer learning models
  2. Enable Training of audio and image classification models from within the app inventor's property pannel instead of visiting external sites
  3. remove other external dependencies like spectrogram generation api and make it work internally

Iā€™d love to hear your ideas or feedback on how we can make this project a success. If you have any thoughts on improving these components, please share them. Your insights are invaluable to me.

3 Likes