The research focuses on the design and construction of an image-based disease classification model based on deep learning techniques. The model aims to classify plants' leaves and stems into one of the five categories: black spots, downy mildew, powdery mildew, healthy, and other diseases. A dataset downloaded from a public sourced datasets and preprocessing techniques involving resizing, color normalization, and augmentation techniques, like rotation and flip, are used to enhance the model's efficiency. The deep model, built from the use of TensorFlow and the Keras API, performs image-based extraction of the images through the use of the convolution layer, enhancing the model's ability to distinguish between different health states in plants. Evaluation results show that the model achieves around 40% confidence when identifying one of the three diseases, indicating reasonable performance with room for future improvement. The model's efficiency in terms of classification measurement confirms the model's reliability in disease prediction. The current research identifies the use of deep models in the future in the field of agriculture to offer scalable and automatable means to disease prediction and monitoring in plants.