Fish disease Detection

• Tech Stack: Python, google colab notebook, MATLAB, Tensorflow, Keras 3 API
• Research Paper: Link
• Github URL: Project Link

A thorough comparison has been conducted on the effectiveness of the top seven pre-trained deep CNN models, including VGG-16, MobileNetV2, InceptionV3, ResNet-50, ResNet-34, EfficientNetB7, ConvNeXtXLarg. Various hyperparameters such as learning rate, batch size, number of epochs, and optimization techniques have been analyzed for their impact. Ultimately, the most optimal model has been identified, providing researchers with a foundation for developing a more efficient CNN-based solution for early detection of fish disease infection.

In order to address the limited and imbalanced nature of the data within publicly accessible datasets, we implemented multi-operation data augmentation techniques, ensuring equal representation of samples from seven different classes.

Our novel approach leverages the collective power of three pre-trained models for deep learning feature extraction, seamlessly integrating them to enhance SVM classification. Our discoveries showcase the transformative capabilities of these methods, setting new standards for performance benchmarks.