How anandchauhan21/pomegranate_disease_prediction works
Project type & purpose (from README + tree)
This is not a traditional “application repo” (no src/, no service code). It’s a research / ML prototype organized as Jupyter notebooks that:
• Trains CNN image classification models (TensorFlow) for:
1) Pomegranate fruit disease prediction
2) Pomegranate leaf disease prediction
• Runs inference/prediction notebooks that likely:
• Load a trained model
• Preprocess an input image (OpenCV)
• Predict a disease class
• Class to a solution/recommendation
• Appears to be intended for Google Colab usage (badge).
What the file tree implies about architecture
Top-level artifacts are notebooks only:
• pomodetectCreateModel_1.ipynb — training pipeline for fruit
• pomodetectPredict.ipynb — inference pipeline for fruit
• LeafdetectCreateModel.ipynb — training pipeline for leaf
• LeafdetectPredict.ipynb — inference pipeline for leaf
• Tcp_spi.ipynb — ambiguous; could be:
• hardware integration experiment (TCP/SPI), edge/IoT capture pipeline, or data transfer
• LICENSE, README.md
This suggests two parallel model tracks (fruit + leaf), each with a create/train notebook and a predict notebook.
What the README adds (stack & behavior)
Key technologies:
• TensorFlow (model definition/training/inference)
• OpenCV (image I/O + preprocessing)
• Colab (execution environment)
Key functional requirements:
• Disease classification from image
• Provide “corresponding solutions” for each disease (so there’s a label → recommendation mapping layer in inference)
“Don’t overthink it”: maximize component separation (specific to this repo)
• “Load dataset” separate from “Preprocess”
• “Define model” separate from “Train”
• “Evaluate” separate from “Export”
• “Load model” separate from “Predict”
• “Predict” separate from “Recommendation mapping”
This gives the most faithful architecture view despite the absence of modular source code.