• Ms. Jyoti Ahirwar
• Mr. Rajneesh Pachouri
• Mr. Anurag Jain

: This thesis presents a novel deep learning framework for facial age estimation using differential regression, which refines initial age predictions by estimating relative age differences between a query facial image and visually similar reference images from a training database. Unlike conventional absolute regression methods that struggle with heterogeneous aging patterns influenced by ethnicity, gender, and lifestyle, our Baseline Age Regressor (BAR) provides an initial estimate, followed by a Differential Age Regressor (DAR) that leverages nearest-neighbor retrieval based on age and embedding similarity to achieve superior accuracy. We introduce an iterative age augmentation scheme modeling BAR error distributions via Kernel Density Estimation, enabling robust reference selection and progressive refinement. Evaluated on MORPH II and CACD datasets using subject-exclusive protocols, our approach attains state-of-the-art Mean Absolute Errors of 2.47 years and 5.27 years, respectively, while analyzing inherent demographic biases.

Facial age estimation, differential regression, deep learning, reference retrieval, age augmentation, bias analysis, MORPH II

"Multi-Task Facial Age and Gender Prediction Using VGG-Face with MTCNN Preprocessing and Robust Error Handling for Unconstrained Indian Faces ", International Journal of Emerging Technologies and Innovative Research (www.jetir.org), ISSN:2349-5162, Vol.13, Issue 1, page no.c442-c449, January-2026, Available :http://www.jetir.org/papers/JETIR2601253.pdf

"Multi-Task Facial Age and Gender Prediction Using VGG-Face with MTCNN Preprocessing and Robust Error Handling for Unconstrained Indian Faces ", International Journal of Emerging Technologies and Innovative Research (www.jetir.org | UGC and issn Approved), ISSN:2349-5162, Vol.13, Issue 1, page no. ppc442-c449, January-2026, Available at : http://www.jetir.org/papers/JETIR2601253.pdf