This paper proposes an approach based on artificial intelligence to open up new prospects for the protection of endangered species in Zakouma National Park, Chad. This paper analyzes a few major algorithm models used in the detection or segmentation process to determine their accuracy when applied to wildlife. An assessment and a benchmarking were carried out on detection and segmentation algorithms such as Faster R-CNN, Mask R-CNN, YOLO V7, YOLO V8, and YOLO NAS. The process of tracking wild animals is handled by the ByteTrack library. A unique ID is assigned to each animal, enabling it to be identified individually after the detection and recognition process. The counting approach derives from the individual animal identification mechanism: when an animal, recognized by YOLO 8, is first detected during the counting process, then the species-specific counter, initialized at 0, is incremented by one unit. The YOLO 8 algorithm yields the most reliable precision rates, and is therefore chosen for animal detection and identification.