R-CNN Based Deep Learning Approach for Counting Animals in the Forest: A Survey

This review paper delves into the pivotal realm of animal classification using images obtained through diverse techniques in forest environments. A robust framework is introduced, employing Transfer Learning (TL) within a Convolutional Neural Network (CNN) and leveraging the power of the Region-based Convolutional Neural Network (R-CNN) model for the construction of an automated animal identification system. This innovative framework is adeptly applied to analyze and identify focal species within captured images, contributing to the advancement of wildlife monitoring technologies. The dataset under scrutiny comprises 6,203 camera trap images featuring 11 distinct species, including Wild pig, Barking deer, Chital, Elephant, Gaur, Hare, Jackal, Jungle cat, Porcupine, Sambhar, and Sloth bear. The inclusion of this diverse set of species ensures the robustness and applicability of the proposed methodology across a broad spectrum of wildlife scenarios. The integration of Transfer Learning withinthe Region-based Convolutional Neural Network (R-CNN) emerges as a crucial element, showcasing outstanding performance in species classification. Notably, the proposed model achieves a remarkable accuracy rate of 96% on the test dataset after a mere 18 epochs, employing a batch size of 32. This breakthrough holds the potential to expedite research outcomes, foster the evolution of more efficient and dependable animal monitoring systems, and consequently, alleviate the time and effort invested by researchers.In line with ethical considerations, the authors maintain anonymity in their contribution, focusing on the significant strides made in the classification andanalysis of camera trap images within the observed site. This paper positions itself as a noteworthy and impactful contribution to the broader field of wildlife research and technology