Credits go to the FastAI team, Naveen Chanakya, and the Flutter team. I integrated a few different tutorials together to form a SaaS pipeline template. This is the code for this video on Youtube by Siraj Raval on building an image classification startup. There are 3 components here; A web API, a model training script, and a mobile app. The code in this repository is for the starter flutter app. Let's go through the 5 step process below. Pull Requests are always welcome!
What is the image classification service you'd like to build? Once you decide, find a related dataset using these tools
- Run this notebook on your local machine or upload and run it to colab. Replace the bear dataset with your own image dataset. It's retraining a 'resnet34' image classification model. This is transfer learning.
- Save the resulting model pkl file to google drive, save the download link.
- Fork this repository.
- Follow the instructions in its README to deploy it to render
- Once deployed, check that it works.
- Then replace line 12 in 'server.py' of the web example with a link to your own classifier pkl file and re-deploy
- Make any cosmetic changes to the front-end inteface that you'd like
- Install Flutter here
- Download this code
- Open it in android studio as a new flutter project
- it will ask you to 'get' all dependencies, say yes and it'll will all be installed automatically
- Replace the default render link in 'main.dart' to the link to your deployed render app
- Notice the 2 functions for signup and login. This is where your stripe and firebase authentication code will be placed
- See this and this