Agriculture is the backbone of our country's economy and plays an important role in our daily lives. However, crown gall and Anthracnose disease in cotton plants caused by disease-causing pathogens such as fungus and bacteria have reduced agriculture in recent years. If these diseases are diagnosed and treated early enough, they can be controlled. However, manually identifying such diseases at the outset for a wide region of land appears to be exceedingly difficult. As a result, the image processing technology is used to test and train several algorithms such as Convolution neural networks (CNN) with pruned and unpruned model, Inception V3, and Resnet 152V2 to determine the accuracy of recognizing diseases in cotton plants. Image processing is used to execute operations on plant leaves and to identify them as damaged or healthy. To recognize disease in plants, convolution neural networks have three layers: convolution layers, pooling layers, and fully connected layers. These three algorithms were compared in order to get better accuracy to protect a plant from diseases and to improve the yield of plants.