I need a complete, reproducible deep-learning pipeline that takes raw leaf images, learns to recognise plant diseases and then serves the prediction through a Streamlit interface. Because I do not yet have the images, the first task is to identify and download a suitable, well-labelled dataset from Kaggle. Feel free to compare a few candidates, but the final choice should give good class balance and enough samples per disease category. Once the data is in place, walk through exploratory data analysis, preprocessing and augmentation inside a Jupyter notebook. From there, build and tune a convolutional neural network (TensorFlow / Keras or PyTorch are both fine) and report the usual metrics plus a confusion matrix so I can judge class-wise performance. When the model is satisfactory, save it and wrap inference in a clean Streamlit app where a user uploads a single image and instantly sees the predicted disease name along with a confidence score. Deliverables • Notebook (.ipynb) with data loading, EDA, preprocessing, training and evaluation • app.py containing the Streamlit interface ready to run with streamlit run app.py • Serialized model file (H5, PT or SavedModel) • requirements.txt listing every library and version used • README explaining setup, training commands and how to launch the web app The project is finished when the environment from requirements.txt, run the notebook end-to-end without edits, and fire up the Streamlit app to classify a new leaf image successfully. And when I am completely satisified with the project