Rice diseases has adverse consequences on crop yield, and the right finding of rice illnesses is the way to stay away from these impacts. In any case, the current sickness analysis techniques for rice are neither precise nor effective, and exceptional hardware is frequently required. In this review, a programmed finding strategy was created and executed in a cell phone application. The strategy was created utilizing profound learning in view of a huge dataset that contained 33,026 pictures of six sorts of rice infections: leaf impact, bogus filth, neck impact, sheath curse, bacterial stripe illness, and earthy colored spot. The center of the strategy was the Ensemble Model in which submodels were coordinated. At long last, the Ensemble Model was approved utilizing a different arrangement of pictures. Results showed that the three best submodels were DenseNet-121, SE-ResNet-50, and ResNeSt-50, as far as a few credits, for example, learning rate, accuracy, review, and illness acknowledgment exactness. In this way, these three submodels were chosen and incorporated in the Ensemble Model. The Ensemble Model limited disarray among the various sorts of infection, diminishing misdiagnosis of the illness. Utilizing the Ensemble Model to analyze six kinds of rice illnesses, a general exactness of 91% was accomplished, which is viewed as sensibly great, thinking about the appearance likenesses in certain sorts of rice infection. The cell phone application permitted the client to utilize the Ensemble Model on the web server through an organization, which was helpful and proficient for the field determination of rice leaf impact, bogus muck, neck impact, sheath scourge, bacterial stripe illness, and earthy colored spot.