Abstract
Rice bowl, that is the name given to Telangana. In this region that has paddy fields spreading as far as one's eye can see, stands quietly above many farmers' livelihoods. In this regard, the paper puts forth a new mobile application that uses Kotlin and TensorFlow Lite to execute real-time detection for multiple rice diseases. Our model, based on an enhanced DenseNet architecture with the addition of SE blocks and depthwise separable convolutions, achieves an accuracy of 98.8%. The same, lightweight version of the model was converted with care into TensorFlow Lite to allow for real-time inference on Android devices, making this state-ofthe-art technology in agriculture available directly to farmers. It's an app that diagnoses instantaneously prevalent diseases like Bacterial Blight, Blast, and Brown Spot in the leaves of rice to help farmers save complete harvests. This paper elaborates the fine details of our architecture of the model, and the finer process of translations into TensorFlow Lite with an in-depth analysis of the mobile app performance in real-world conditions on various agricultural landscapes in Telangana.