EDUNET FOUNDATION I SHELL I ARTIFICIAL INTELLIGENCE I 4-WEEKS VIRTUAL INTERNSHIP
Link to Dataset - https://drive.google.com/file/d/1Hq8AeXE0HGtphryPUnTA2-FoNEGjmKr_/view?usp=sharing
Link to .h files - https://drive.google.com/drive/folders/1MzlkR5kSpjOsRb2phGU0ScANEOi19kvH?usp=sharing
Link to Colab Notebook - https://colab.research.google.com/drive/1jSDSLsDlPo_Wbb2knjSQk8bnBMLHY4Dm?usp=sharing
WEEK-2 - I mounted the dataset directly from Google Drive into Google Colab, ensuring seamless and persistent access. I analyzed image dimensions to identify inconsistencies and detected duplicate images using MD5 hashing, which helped clean the dataset for better model performance. Additionally, I applied real-time data augmentation techniques such as rotation, zoom, shear, and horizontal flipping using ImageDataGenerator, along with proper training-validation splits. These steps collectively improved data quality and prepared the dataset for robust model training.
Link to Colab Notebook - https://colab.research.google.com/drive/1jSDSLsDlPo_Wbb2knjSQk8bnBMLHY4Dm?usp=sharing
Earlier i compared the three models and at first all the three models were attaining the accuracy ranging between 10 - 30 %, lol!
Now my model is 98% appx. accurate!
Link to Kaggle Notebook - https://www.kaggle.com/code/divijaarrorra/efficient-net
