akhichit2008/bolt-mini

A Mini Image Classifier written in python using tensorflow module and is trained on multiple datasets like "intel image dataset" among others. Note: This project is under development and is presently trained only on the open source intel dataset

Hi there! I'm AKHILESH JC 👋

Welcome to my GitHub profile! I'm a passionate high school student with a keen interest in Artificial Intelligence (AI) and Machine Learning (ML). I love exploring the vast world of AI and using my skills to create projects that simplify tasks and make a positive impact on people's lives.

Get in Touch

I'm always excited to connect with fellow AI enthusiasts, developers, and researchers. Feel free to reach out to me through the following channels:

• 📧 Email: [email protected]
• 🐦 Twitter: @akhileshjc2008
• 😄 Fun Fact: I love discussing the latest AI trends, so don't hesitate to strike up a conversation!

bolt-mini's Issues

Overfitting in base CNN model

the base model located in notebooks/bolt-mini-base.ipynb seems to be overfitted as the training loss is very low but validation loss is sky rocketing. Similar situation goes on with training accuracy and validation accuracy.

Note : The present base model is currently trained only on a subset of the intel dataset specifically only "seg_train" directory has been taken into account and further updates will be made to accommodate remaining data and also include multiple other datasets through a simple pipeline that will merge multiples datasets together for preprocessing.

Possible Fixes :-
1. Increase size of dataset
2. Add keras regularizers (l1,l2)
3. Use data augmentation [Optional]