Geethanjali College of Engineering and Technology
Department of CSE(AI&ML)
PROJECT EXECUTION PROCEDURE DOCUMENT
Title of the Project: Deep Agro
Team Details:
1. Acsah Meduri-21R11A66E5
2. Vaishnavi Taduri-22R15A6619
3. Susheel Vedururu-21R11A66J4
Step-by-Step Execution of a Project:
Step 1: Open Jupyter Notebook and Load Project
Install Jupyter Notebook (if not installed):
pip install notebook
Launch Jupyter Notebook:
jupyter notebook
This will open a new tab in your web browser with the Jupyter interface.
Open an Existing Project:
Navigate to the folder where your `.ipynb` file is located.
Click on the notebook file to open it.
Verify the Kernel:
Ensure that the correct Python environment is selected.
If needed, change the kernel via 'Kernel' → 'Change Kernel'.
Step 2: Execute Notebook Cells
Run Cells One by One: Click on a cell and press `Shift + Enter` to execute it.
Run All Cells: Click on 'Run' → 'Run All' to execute all cells sequentially.
Restart and Run All Cells: If there are errors or changes, go to 'Kernel' → 'Restart &
Run All'.
Check for Errors: If an error occurs, debug and rerun the affected cells.
Step 3: Data Preprocessing
Collect Data: Gather images, text, or numerical datasets.
Data Cleaning: Remove duplicates, handle missing values, normalize/standardize
values.
Data Splitting: Split data into training, validation, and test sets.
Augmentation (Optional for images): Apply transformations like rotation, flipping,
cropping.
Convert to Tensors: Convert images or text into tensors using NumPy or TensorFlow.
Step 4: Model Design
Choose Model Type: CNN, RNN/LSTM, Transformers, MLP.
Define Layers: Input Layer, Hidden Layers, Output Layer.
Initialize Weights and Biases: Use Xavier or He initialization.
Step 5: Model Compilation
Choose Loss Function: Cross-entropy loss for classification, MSE for regression.
Select Optimizer: SGD, Adam, RMSProp.
Define Metrics: Accuracy, RMSE.
Step 6: Model Training
Forward Propagation: Pass input through layers.
Loss Calculation: Compute how far predictions are from actual values.
Backward Propagation: Compute gradients and update weights.
Repeat for Multiple Epochs: Iterate over dataset multiple times.
Step 7: Model Evaluation
Use Validation Data: Measure accuracy, precision, recall, F1-score.
Plot Loss & Accuracy Curves.
Hyperparameter Tuning: Adjust learning rate, batch size, number of layers.
Step 8: Model Testing
Compute Performance Metrics: Compare test accuracy with validation accuracy.
Check for Overfitting: Identify if test performance drops significantly.
Step 9: Model Deployment
Convert Model Format: Use TensorFlow SavedModel, ONNX, or TorchScript.
Optimize for Inference: Use quantization, pruning, hardware accelerations.
Deploy via API: Use Flask, FastAPI, AWS, GCP, Azure.
Monitor Performance: Use logging, A/B testing, feedback loops.
Signature of the Guide