The researchers used a deep learning method to detect the growth period of deciduous broad-leaved forests from single PhenoCam images. They achieved high accuracy in identifying the growing regions and periods, with up to 10.2% accuracy on a daily scale and 69% accuracy on an 8-day scale. When divided into four categories, the accuracy of identifying vegetation reached 98.2%. This method improves upon traditional techniques by providing faster and more precise results, contributing to the study of vegetation growth and phenology.