Over 25% of the world population faces anemia as a global health problem which creates severe complications carrying negative effects on child brain development and adult work capacity and posing life-threatening risks during pregnancy. The current blood tests, which use complete blood count (CBC) and hemoglobin analysis, require blood collection and specialized equipment and medical facilities, thus preventing their use in underserved regions. Modern medical imaging, together with artificial intelligence (AI) technology, creates new possibilities for screening anemia without invasive procedures when focusing on evaluating ocular conjunctival pallor as a validated clinical indication of low hemoglobin levels. Visually supervised deep learning (DL) architectures learn automated feature extraction from primary raw images maintaining spatial relationships which goes beyond traditional machine learning features based on human intervention. A hybrid Convolutional Neural Network (CNN)-Vision Transformer (ViT) model developed in this research addresses three major issues by identifying microvascular patterns while measuring pallor distribution extent and linking detected features to hemoglobin limits. Using datasets annotated by hematologists, our system shows how computer vision technology can perform clinical pallor evaluation using lab-quality results, without requiring blood draws, which creates an important development for community health workers along with telemedicine programs.