Automated Tiger Detection System Using Faster R-CNN and MobileNet Architecture

Abstract

The growing importance of wildlife conservation is highlighted by the declining numbers of endangered species like the tiger. It is essential that we develop tools for the precise and automated monitoring of tiger populations in the wild for conservation purposes. Current techniques for population monitoring rely on laborious manual checks, and the use of camera traps, which are susceptible to random human error. The increased use of deep learning techniques has led to increased interest in the automation of animal detection through the use of object detection models. Unfortunately, in the case of tiger detection, there are a number of unique challenges, such as occlusion, changes in the light, the presence of dense undergrowth, and the small size of the target. We propose a detection system that is efficient in terms of tiger detection and based on deep learning techniques, particularly the Faster R-CNN model with a MobileNet backbone. MobileNet supports a lightweight architecture for real-time use, while Faster R-CNN is able to achieve a high detection rate because of its regional proposal approach. These values represent the high levels of diverse safety detection performance cited in the methodology. In closing, the proposed Faster R-CNN with MobileNet backbone architecture works as a strong detector and value-adding MobileNet backbone architecture. The system proposed has a place in the systems used in the monitoring of wildlife, providing the value of detection systems in the field for tracking and in the protection of wildlife.