Machwala: Android Application for Local Fish Detection and Information Retrieval in Bangladesh

This research involves the development of an Android application tailored for fish detection in local contexts, specifically focused on Bangladesh. Employing Convolutional Neural Network (CNN) image classification, three variations of the VGG16 model were explored, each incorporating distinct classifiers. After rigorous evaluation, the VGG16 model with MLP classifiers emerged as the preferred choice, boasting a commendable 95% accuracy within a moderate training time. This model was seamlessly integrated into an Android app crafted using Java and Android Studio. The application follows a user-friendly flow: capturing and preprocessing user-provided images and applying the MLP-based VGG16 model for species identification. Beyond mere classification, the app enriches the user experience by elucidating consumption benefits, potential drawbacks, and approximate market pricing for the identified fish species. Ambitious future plans encompass enhancing user interaction, expanding the dataset to accommodate more fish species, integrating reinforcement learning capabilities, and connecting with online fish market APIs to predict real-time fish prices.