Creating Alert Messages Based on Wild Animal Activity Detection

Abstract

This study seeks to address the serious issue of animal attacks on the rural populace and forestry workers by looking into a viable method of surveillance. We present a Hybrid Visual Geometry Group-19 (VGG-19) and Bidirectional Long ShortTerm Memory (Bi-LSTM) network to identify the types of animals, track their movements and provide a real-time location indicator for safety alerts in the forest area. The suggested model is comprised of two parts: VGG-19 for feature extraction and BiLSTM for sequence learning, which is more accurate than the traditional surveillance techniques, and can detect the animals and their movement pattern. Moreover, a combination of predictions from a set of different models, called ensemble method, is used to enhance robustness and accuracy. Moreover, we obtain 100% accuracy after studying some methods like CNN+BiGRU to enhance the performance. In addition, the system has a userfriendly front end developed by Flask framework which enables testing of the system by users with authentication features. This work provides a viable method for reducing the chances of animal attack by exploiting advanced DL algorithms and user-centric design for effective safety monitoring in rural and forestry settings