Mobile Menagerie: Evaluating and Implementing CNNs for Real-Time Animal Species Recognition on Smart Devices

An Animal Detector is a cutting-edge mobile application that revolutionizes real-time animal species classification using TensorFlow Lite (TFLite), Flutter, and sophisticated machine learning techniques. With over 28,000 medium-quality animal photos carefully arranged into ten different categories, the program boasts a sizable dataset. It can classify a vast variety of creatures, from cherished domestic pets like dogs and cats to fascinating wildlife like elephants and butterflies. Thorough testing has improved Animal Detector's accuracy to an astounding 98%. This includes experimenting with different image recognition networks, most notably Google Inception. One particularly useful feature allows users to take photographs of animals in real-time with their mobile devices' cameras, which helps researchers quickly identify and record wildlife while they are out in the field. This multifunctional program can categorize images from various sources, so users can classify dynamic snapshots and animal photos from their device's library. Even though the convolutional neural network (CNN)-based model initially obtained an accuracy rate of 82%, the app is constantly evolving and adding new features that confirm its value as a significant resource for the wildlife study community. The development and capabilities of the Animal Detector application are described in this study, highlighting its critical role in developing real-time animal species classification, aiding in conservation efforts, and facilitating nature exploration wherever it may be.