Projects

MACHINE LEARNING-BASED DIGIT RECOGNITION: BEYOND DEEP LEARNING

•Developed a pipeline to evaluate SIFT and SURF feature extraction techniques on the MNIST dataset.

•Implemented SIFT and SURF to extract keypoints and descriptors from images.

•Trained Random Forest Classifiers on individual and combined feature sets, with and without PCA.

•Conducted experiments with varying training sample sizes from 50 per class to 300 per class, visualized results, and documented findings.

CUSTOM IMAGE ROTATION AND CROPPING WITHOUT DEPENDENCIES 

• Precision Image Rotation: Rotate images with the option to maintain original dimensions or showcase the full rotated view.

• Black Pixel Detection: Detection of black pixels to calculate the unintended rotation angle in images.

• Automatic Correction: Automatically correct the image orientation based on detected angles, ensuring perfect alignment.

• Efficient Cropping: Removes unwanted black borders after correction, leaving only the content you care about.

• Scalable Processing: Downscales images for quick analysis and then upscales after correction, preserving the original quality.