Image Detection and Recognition of different species of animals using Deep Learning
R. Shanthakumari, C. Nalini, S. Vinothkumar, B. Govindaraj, S. Dharani, S. Chindhana
Abstract
Deep Learning has grown in popularity as a method for solving computer vision difficulties. Deep learning models are becoming increasingly efficient in solving complex real-world issues, as evidenced by numerous recent articles. However, training the networks slowly takes time and money for certain real-world situations that are difficult to handle with deep learning. It is vital to find ways to collect data rapidly so that it can use it in real-time applications. With a growing awareness of the need for habitat conservation, researchers are examining animal density, presence, and absence to determine the visibility of endangered species. Animal detection tools are required to estimate and monitor the quantity of animals in a specific area. This can be considered a subset of environmental monitoring. Camera-based technologies, as well as acoustic and seismic measures, can all be used to detect animals. Because cameras are a helpful tool for monitoring wildlife behaviour and assessing ecosystems, they have a long history of use. Animals, on the other hand, can move quickly and conceal themselves to avoid being spotted. The YOLOV5 model could be used to detect several types of animals in the forest. From photos, real-time camera feeds, and recorded videos, the YOLOV5 model can recognize horses, sheep, cows, elephants, bears, zebras, and giraffes etc. It also has the ability to detect birds. Objects can be detected in real time by YOLOV5 using a camera feed attached to the CPU via USB connection. The advantages of the YOLOv5 network’s is lightweight so it reduces the deployment cost of the identification model. Furthermore, animal detection may be required for a variety of purposes, including livestock management, wildlife management, pest control, and security.