Urban Lawns State Identification Method Based on Computer Vision

The article discusses an approach to assessing the condition of urban lawns based on computer vision. The assessment of the condition of urban lawns is based on the identification of the composition of plant species growing on the lawn, since the species composition can characterize the state of the soil and growing conditions, as well as the characteristics of some external factors. Traditionally, the assessment of the condition of urban lawns is carried out by the method of full-scale research and requires considerable time and resources of specialists. The paper investigates the hypothesis of the possibility of automating this process using computer vision methods. To test this hypothesis, we used the YOLOv5 neural network for object recognition. The model was fine-tuned on a carefully curated dataset of lawn images that had been annotated with rectangles. The study revealed the shortcomings of the model, the most recognizable classes, as well as opportunities for development and improvement of the model. The proposed approach makes it possible to develop a large service for analyzing the state of urban green spaces.