Welcome to the Computer Vision Projects repository. This collection features a series of projects that delve into various advanced techniques in Computer Vision. Each project is a demonstration of cutting-edge methodologies applied to solve specific tasks in the realm of image processing and analysis.
- 🌉➡️🌆 Image-Image-Translation Using CycleGANs
- 🕵️♂️📷 Object Detection Using YOLO-NAS
- 📸 Panoramic Image Stitching Project
- 🔍 Histogram Equalization Assignment
This project explores the use of Cycle-Consistent Generative Adversarial Networks (CycleGANs) for unpaired image-to-image translation. The focus is on translating images from one domain to another without the need for paired training examples.
- Unpaired Image Translation: Translate images between different domains without paired training data.
- Cycle Consistency Loss: Ensures that translated images can be mapped back to the original domain.
- Adversarial Training: Uses GANs to generate realistic images in the target domain.
This project demonstrates how to build a computer vision interface using Streamlit in Python for real-time object detection. It utilizes the YOLO NAS (You Only Look Once - Neural Architecture Search) object detection model, which is known for its efficiency and accuracy.
- Real-Time Object Detection: Perform object detection on images and videos in real-time.
- Streamlit Interface: User-friendly web interface for uploading images and videos, and viewing detection results.
- High Accuracy: Utilizes YOLO NAS for precise object detection with optimized network architecture.
This project explores the process of combining multiple photographic images with overlapping fields of view to produce a segmented panorama or high-resolution image using homography.
- Feature Detection using SIFT: Identifies distinctive points or regions in the images.
- Feature Matching: Matches features between images using brute force and K-Nearest Neighbors (KNN).
- Homography and Image Blending: Aligns and blends multiple images to create a seamless panorama.
Histogram Equalization (HE) is a fundamental image processing technique used to enhance the contrast of images. This project evaluates several HE methods, including Global HE, Adaptive HE, and Contrast-Limited Adaptive Histogram Equalization (CLAHE).
- Global Histogram Equalization: Enhances the global contrast of an image.
- Adaptive Histogram Equalization: Enhances local contrast by dividing the image into smaller regions.
- Contrast-Limited Adaptive Histogram Equalization (CLAHE): Limits the contrast enhancement to avoid noise amplification.
The Computer Vision Projects repository demonstrates the application of advanced computer vision techniques across various domains. Each project showcases the potential and versatility of modern computer vision frameworks in solving complex image processing tasks.
For detailed information and setup instructions, please refer to the individual README files for each project. If you have any questions or feedback, feel free to reach out to the project contributors.