Vitamin Deficiency&Retinal Analysis Image Dataset
Description
Vitamin Deficiency and Retinal Analysis Image Dataset Medical image dataset for nutritional deficiency detection and retinal vessel analysis. Contains six classes: Vitamin A, Vitamin B, Vitamin C, Vitamin D, Vitamin E, and Retina Blood Vessel. Images are organized in class-specific folders within the main dataset directory. The dataset includes several hundred images across the six classes, with varying counts per class. Images are in PNG, JPG, and JPEG formats (case-insensitive). All images are converted to RGB during loading to ensure consistent color representation. Preprocessing standardizes images to 224x224 pixels. Normalization uses ImageNet statistics (mean [0.485, 0.456, 0.406] and standard deviation [0.229, 0.224, 0.225]) to align with standard computer vision practices. Data augmentation is applied during training: random resized crop from 256x256 to 224x224, random horizontal flip (50% probability), random rotation up to 15 degrees, color jitter (brightness, contrast, saturation, hue), and random affine translation. These augmentations improve generalization and robustness. The dataset structure supports train/validation/test splits, typically using stratified sampling to maintain class distribution. Mask directories and mask-related files are excluded during loading. Images are loaded using PIL/Pillow with error handling that provides placeholder images if loading fails. This dataset supports medical image classification tasks focused on nutritional deficiencies and retinal vessel patterns, enabling automated analysis of clinical images for diagnostic assistance and research.
Files
Steps to reproduce
dataset/ (with subdirectories: Vitamin A/, Vitamin B/, Vitamin C/, Vitamin D/, Vitamin E/, Retina Blood Vessel/)