Kannada Text Detection in Video Frames
Detecting Kannada text in real-world video frames
A computer-vision pipeline exploring text-region detection in Kannada news-video frames under blur, low resolution, and changing scene conditions.
What it does
- Extracts frames from Kannada news videos using OpenCV for dataset preparation.
- Explores text-region proposals and YOLOv8-based detection, with a focus on Kannada script in real-world footage.
- Includes annotation, dataset organization, augmentation, training, and inference workflows.
- The project is being developed to train and evaluate a custom detector rather than rely on a pretrained OCR engine for the core detection task.