An Analysis of Advanced Automated Tree Detection and Species Classification Utilizing High-Resolution Imaging and YOLO-NAS

Abstract

Trees play a crucial role in sustaining our planet by producing oxygen, storing carbon, and offering habitats for wildlife. Artificial intelligence enhances their management by automating the detection of trees and monitoring their health, leading to more intelligent and efficient conservation efforts. Environmental preservation and the prompt and precise combat of climate change are aided by this technology. For monitoring purpose UAV’s/ mobile camera/CCTV camera with AI enabled DL based models can achieve the objective discussed. In this study, we investigated DL based YOLO-NAS model's capacity to recognize whole trees in digital photos taken with high-definition or mobile cameras. According to our results, YOLO-NAS successfully detects single trees with high confidence scores and precise bounding boxes. A diverse set of photos from Google and real-time photos taken with Android phones were used to evaluate this strategy. YOLO-NAS recorded mean Average Precision (mAP) around 87.2%, Precision around 88.0%, and Recall of around 80.2% when compared to YOLOv8. However, with a mAP of 88.0%, Precision of 86.9%, and Recall of 85.1%, YOLOv8 fared better than YOLO-NAS. The two models were similarly powerful, with YOLOv8 providing superior recall and YOLO-NAS demonstrating superior precision.