Computer Vision Course - 7th Semester

Student: Muhammad Abdullah Awan
Roll Number: 2022-SE-08
University: The University of Azad Jammu and Kashmir
Semester: 7th
Date: February 2026

---

📁 Repository Structure

This repository contains all coursework for the Computer Vision course, organized into the following folders:

Computer_Vision/
├── Assignment/              # Individual Assignment Work
├── Quiz/                    # Quiz 1 (Individual)
├── Quiz2/                   # Quiz 2 (Individual)
└── computer_vision_project/ # Team Project

---

📂 Folder Descriptions

1. 📝 Assignment (Individual Work)

Student: Muhammad Abdullah Awan (2022-SE-08)

Contains individual assignment work focused on hybrid image creation and frequency domain image processing.

Contents:

• hybrid_image_assignment.ipynb - Hybrid image generation using frequency domain techniques

Topics Covered:

• Gaussian filtering
• Frequency domain processing
• Low-pass and high-pass filtering
• Hybrid image creation

---

2. 📋 Quiz (Individual Work)

Student: Muhammad Abdullah Awan (2022-SE-08)

Contains Quiz 1 work on dynamic slip cropping and processing.

Contents:

• Dynamic_Slip_Cropping.ipynb - Automated slip detection and cropping
• cropped_slips/ - Output folder for processed images

Topics Covered:

• Image preprocessing
• Object detection
• Dynamic cropping techniques
• Batch image processing

---

3. 📋 Quiz2 (Individual Work)

Student: Muhammad Abdullah Awan (2022-SE-08)

Contains Quiz 2 work on dollar bill value detection using deep learning.

Contents:

• Dollar_Bill_Detection_Quiz2.ipynb - CNN-based dollar bill classifier

Topics Covered:

• Train/Test data splitting
• CNN architecture design
• Transfer learning with EfficientNetB0
• Image classification
• Model evaluation and accuracy testing

Key Requirements Met:

• ✅ Created separate test folder with same structure as training
• ✅ Ensured no overlap between train and test datasets
• ✅ Trained CNN model for multi-class classification
• ✅ Tested accuracy on independent test data

---

4. 🎯 computer_vision_project (Team Project)

Team Members:

• Muhammad Abdullah Awan (2022-SE-08)
• Umair Imtiza Khokhar (2022-SE-18)
• Awais Ahmed Abbasic (2022-SE-29)

Project: ECG Image Digitization - Kaggle Competition

Contains team project work for the PhysioNet ECG Image Digitization Challenge, converting ECG images to digital time-series signals.

Contents:

• README.md - Detailed project documentation
• ecg_training.ipynb - Approach 1: Custom CNN training
• ecg-testing.ipynb - Approach 1: Testing and submission
• open-ecg-digitizerv6.ipynb - Approach 2: Multi-stage pipeline (Best Performance)

Project Overview:

• Competition: PhysioNet ECG Image Digitization on Kaggle
• Task: Convert 12-lead ECG images to digital waveforms
• Evaluation Metric: Signal-to-Noise Ratio (SNR)

Approaches Compared:

Approach	Kaggle Score	Description
Approach 1: Custom CNN	0.08	Simple end-to-end learning
Approach 2: Multi-stage Pipeline	17.1	Pre-trained models + domain knowledge

Key Learnings:

• Multi-stage processing outperforms end-to-end learning for complex tasks
• Domain knowledge integration is crucial for medical imaging
• Transfer learning significantly improves performance
• Task decomposition (normalization → grid detection → signal extraction) is effective

---

🎓 Course Topics Covered

Image Processing Fundamentals

• Frequency domain analysis
• Gaussian filtering
• Image transformations
• Preprocessing techniques

Computer Vision Techniques

• Object detection and cropping
• Image segmentation
• Feature extraction
• Grid and keypoint detection

Deep Learning for Vision

• Convolutional Neural Networks (CNN)
• Transfer learning (EfficientNetB0, MobileNetV2, ResNet)
• Data augmentation strategies
• Model training and optimization
• Multi-stage pipeline architectures

Model Evaluation

• Train/Test splitting strategies
• Accuracy metrics
• Confusion matrices
• Classification reports
• SNR (Signal-to-Noise Ratio) for signal processing

---

📊 Work Classification

Work Type	Project/Quiz	Student(s)	Folder
Individual	Assignment	Abdullah (Roll 8)	Assignment/
Individual	Quiz 1	Abdullah (Roll 8)	Quiz/
Individual	Quiz 2	Abdullah (Roll 8)	Quiz2/
Team	Project	Abdullah, Umair, Awais	computer_vision_project/

---

🛠️ Technologies Used

Programming Languages

• Python 3.x

Libraries & Frameworks

• Deep Learning: TensorFlow, Keras, PyTorch
• Computer Vision: OpenCV, PIL/Pillow
• Scientific Computing: NumPy, SciPy
• Data Manipulation: Pandas
• Visualization: Matplotlib, Seaborn
• Image Processing: scikit-image, Albumentations
• Pre-trained Models: timm (PyTorch Image Models)

Development Environment

• Google Colab (for GPU acceleration)
• Jupyter Notebook
• VS Code

---

📈 Performance Highlights

Quiz 2: Dollar Bill Detection

• Successfully implemented train/test splitting
• Achieved high accuracy using transfer learning
• Proper data augmentation and model evaluation

Team Project: ECG Digitization

• Best Kaggle Score: 17.1 SNR
• 213x improvement over naive approach
• Demonstrated importance of domain-specific design
• Multi-stage pipeline outperformed end-to-end learning

Assignment: Hybrid Images

• Successful frequency domain manipulation
• Proper implementation of Gaussian filters
• Created perceptually interesting hybrid images

---

📝 Notes

• Individual Work: Assignments and quizzes are completed independently by Muhammad Abdullah Awan
• Team Work: The ECG project is collaborative team effort
• Code Quality: All notebooks include detailed comments and explanations
• Reproducibility: Random seeds set for consistent results
• Documentation: Comprehensive README files in project folders

---

📞 Contact

Muhammad Abdullah Awan
Roll Number: 2022-SE-08
Course: Computer Vision (7th Semester)
University: The University of Azad Jammu and Kashmir

---

📄 License

Educational coursework for university requirements.

---

Last Updated: February 2, 2026