Class-Gen is a generation classification project majorly for 4 generations of people.
- The Gen z
- The Gen x
- The Baby boomers
- The Millenials
This project is for fun and it is just to learn how fine-tuning pre-trained models, image augmentation and preprocessing works. I decided to go ahead with this project because there's no exact solution for it yet and we just have to keep trying till the neural networks learn perfectly.
Once, a validation accuracy of 75% has been reached, I am going to deploy and release the web application, for people to play around and give feedback.
Here's what I have done so far:
- Scraped the web for images of 4 generations.
- Data cleaning and preprocessing.
- Image Augmentation
- Used inception v3 on the model(tried resnet too, but it hasn't worked this fine)
- Downloaded the .pkl format of the model with 64% val accuracy
Here are other ways you can contribute to this project.
- Other type of pre-trained model - multimodal llm's might be a way to go
- Working on this project using pytorch or other deep learning frameworks
Please note that you can contribute in other ways too, just make sure to raise an issue before opening a pull request.
- Python.
- Selenium.
- Computer vision.
You can always reach out to me via email at mardiyyahodu@gmail.com or via Twitter @hemhemoh for further clarifications