Abstract: Detecting and keeping tabs on wild animals is super important for protecting wildlife preventing human-wildlife conflicts and keeping an eye on our forests traditional methods like basic CNN object detection and machine learning can really struggle with busy backgrounds tracking movement and working in real-time to tackle these issues our project introduces a hybrid deep neural learning model that brings together VGG-19, Bi-LSTM and CNN-GRU to accurately and efficiently identify wildlife we use VGG-19 to pull out key spatial features making it easier to recognize different species Bi- LSTM helps us capture temporal relationships allowing us to analyze movement patterns across video sequences on top of that we employ CNN-GRU to optimize computation performance enabling smooth real-time processing while keeping accuracy high our model is trained on large datasets that cover a variety of animal types and environmental conditions. Our test results confirm that VGG- 19, Bi- LSTM and CNN-GRU model nails a classification accuracy of 98.2 greatly beating out traditional CNN methods and popular detection systems like yolo and faster r-cnn this system with its low false positives and negatives is ready for real-world uses in forest monitoring and wildlife conservation this study really emphasizes how deep learning can step up wildlife detection tracking and classification finally boosting safety and conservation efforts.
Index Terms: VGG-19, Bi-LSTM, CNN, CNN-GRU, Object Detection, Deep Learning, Accuracy, Spatial features, Human- Wildlife Conflict, Environmental conditions.
Copyright Form