In this project, I conducted a comprehensive comparison between a scratch CNN model and fine-tuned models, evaluating their performance on the Intel Image Classification dataset. These approaches were analyzed in terms of accuracy, training efficiency, and suitability for the task, providing valuable insights into model selection and optimization strategies.
pip install -r requirements.txt
Size : 363 MBs
Link : https://www.kaggle.com/datasets/puneet6060/intel-image-classification
Classes: Buildings, Forest, Glacier, Mountain, Sea, Street
Folders: Train, Test and Prediction
Total Images: 24,000+
Convolution Neural Network - IntelNet
- Finetuned: MobileNet-V2
- Finetuned: Wide ResNet-50
- Finetuned: VGG-16