Cotton, accounting for approximately 25 percent of global fiber usage, is significant among textile fibers worldwide. However, the water resources are currently insufficient to provide full irrigation in the arid or semi-arid areas for cotton. Therefore, finding effective methods to optimize irrigation water use is essential. This workshop aims to classify cotton water stress using convolutional neural networks. The high-resolution cotton RGB image was collected by a DJI Phantom 4 at Lubbock, Texas. The study included three replications to evaluate four irrigation treatments. The research results demonstrated that the CNN model successfully classified the cotton water stress with an overall accuracy of 91%.