While going through Part 1 of this course, I faced some difficulty in sending images from my phone to my desktop conveniently. Another problem was getting real-time predictions for new images. Also, I wasn’t happy with bad quality pics taken on a webcam. I wanted a better quality image source. This meant using my phone. But using my phone meant transferring the pic to my server through some means (cable, google drive etc) and then testing my model on the image(which is a big hassle).
I ended up creating a tool for myself to do the following
- Get quick predictions from the server on pics taken from the phone.
- Send images easily from phone to the server.
- Trigger training and pass training parameters from the app.
- Switch between multiple servers easily
I built the app for personal use. It helped me curate a lot of custom datasets as well as test my model’s performance on new images easily which ultimately made the course a lot more fun for me. I thought I would share the app here for others who might find it useful too.
Currently, the app is available on Expo. This means that it will only work on Android phones. Python version for server code is Python 3. The app and the server code is open source. The app was only ever intended for personal use. As such, I haven’t put in a lot of effort into making it high quality or secure but for the purposes of learning, it works, I suppose. Feel free to improve it and use it as you wish
Link to server code and usage/installation instructions - https://github.com/prasannals/image_source_server
Link to app code - https://github.com/prasannals/image-source
Server setup walkthrough video - https://youtu.be/aglR4Y3AdWY
Prediction Demo - https://youtu.be/Fe8uBVMcH8U
Please let me know if you’re facing any trouble with the usage of the app. I will try to reply and help ASAP