🖼️ Task 4: Image Classification with CNN
An intermediate-level Deep Learning project developed during my internship at VedGrow. This project implements a Convolutional Neural Network (CNN) using TensorFlow & Keras to classify handwritten digits from the classic MNIST dataset.
🏢 Project Overview
The primary goal of this project is to build a highly accurate computer vision pipeline capable of recognizing and categorizing handwritten digits (0-9) from image pixel matrices. The engine optimizes mathematical parameter layers via deep convolutional networks to recognize abstract structural features.
🚀 Key Features Implemented:
- 📥 Instant Dataset Ingestion: Configured integrated loader matrices to download and stream MNIST array sets seamlessly within seconds.
- ⚙️ Data Augmentation Filters: Implemented sequential transformation pipelines mapping random rotations and zoom ratios to handle variations and prevent model overfitting.
- 🧠 CNN Layer Architecture: Structured an optimal deep neural network layout combining multiple Convolutional layers, Pooling components, Flatten operations, and a 10-node Softmax classification layer.
- 📈 Performance Logs Verification: Visualized network metric tracking baselines by plotting separate standalone training vs. validation loss/accuracy curves.
- 🎯 Custom Predictive Inference: Integrated verification sandbox workflows that select unseen images and extract correct classifications alongside computed probability confidence indicators.
🛠️ Tech Stack & Frameworks
- Language: Python 3.x
- Deep Learning Framework: TensorFlow 2.x / Keras
- Matrix Operations: NumPy
- Data Visualization: Matplotlib, Seaborn
📂 Project Structure
📂 image_classification_cnn/
├── 📄 cnn_classifier.ipynb # Structured Jupyter Notebook with 6 Core Sections
├── 📄 cnn_model_accuracy_curves.png # Saved visualization chart for model accuracy trends
├── 📄 cnn_model_loss_curves.png # Saved visualization chart for loss minimization
└── 📄 cnn_live_custom_prediction_test.png # Output proof plot verifying successful live inference
⚙️ Setup and Execution
- Open the project root folder space inside your active system command terminal.
- Install the target pipeline package criteria dependencies using the inline configuration:
pip install tensorflow numpy matplotlib seaborn
- Run the notebook sections sequentially from
Cell 1 through Cell 6 inside your VS Code Jupyter environment.
👤 Author