This project implements an object classification model that leverages both traditional image processing techniques and deep learning (YOLO-v8) to identify and classify objects in raw images. The model is designed to efficiently process raw image data and generate accurate predictions for various objects in real-world scenes.

• Image Preprocessing: Applied multiple image filtering techniques to enhance raw images.
  • MedianBlur
  • Bilateral filtering
  • Gaussian Blur
  • Fourier-based filters
• Object Detection and Classification: Trained a YOLO-v8 model for robust object classification and localization.
• Accurate Predictions: Achieved high precision and recall in identifying and classifying objects from images.

• Deep Learning Framework: YOLO-v8
• Image Processing: OpenCV for image filtering and enhancement
• Programming Language: Python
• Other Libraries:
  • NumPy
  • pandas
  • Matplotlib for visualizations
  • Ultralytics for YOLO

• Image Preprocessing:
  • Applied MedianBlur, Bilateral, and Gaussian filters to reduce noise and improve image quality.
  • Fourier-based filtering for frequency domain enhancements.
• YOLO-v8:
  • Utilized the YOLO-v8 architecture for real-time object detection.
  • Configured bounding boxes and class predictions using the YOLO head layer.

• The model was trained on a dataset of raw images containing objects from various categories. Each image was labeled with bounding boxes and class labels to facilitate training.

• Preprocessing: The raw images underwent several transformations including noise reduction and edge enhancement, which improved the model’s ability to detect objects in noisy and low-quality images.

• The YOLO-v8 model successfully generated accurate object predictions with high precision and recall.
• The application of image filtering techniques prior to model training led to improvements in detection accuracy, especially for noisy or low-contrast images.