Mandi, Himachal Pradesh
Researchers at the Indian Institute of Technology (IIT) Mandi are developing an artificial intelligence (AI)-based system to assist forensic experts in identifying unidentified human remains by analysing skull structures and matching them with facial images. The technology could help investigators narrow down potential identities in cases where conventional identification methods face significant challenges.
According to a report published by The Tribune, the research addresses an important challenge in forensic investigations in India, where approximately 50,000 bodies reportedly remain unidentified each year, based on data cited from the Unidentified and Missing Persons Database maintained by AIIMS.
AI-Powered Approach to Forensic Identification
The research is being conducted by the Visual Intelligence and Machine Learning (VIML) Group at IIT Mandi’s School of Computing and Electrical Engineering. PhD research scholar Ravi Shankar Prasad is undertaking the work under the supervision of Assistant Professor Dinesh Singh.
The project, which began in August 2023, explores how AI can learn the relationship between human skull morphology and facial appearance. By analysing datasets containing skull information and corresponding facial images, the model identifies patterns that may assist in forensic identification.
The system is designed to perform two principal functions. First, it can compare skull-related information with facial images in a database and generate a shortlist of potential matches. Second, when photographs of an unidentified individual are unavailable, it can use skull information to generate a probable facial representation.
These capabilities could support investigations involving decomposed, skeletonised or otherwise unidentified human remains, particularly when investigators have limited information about the deceased.
Potential Advantages for Forensic Investigations
Traditional facial reconstruction can involve specialised forensic expertise and techniques such as clay modelling. An AI-assisted approach could make certain stages of the process more systematic and potentially reduce the time required to generate investigative leads.
The technology may have applications in missing-person investigations, unidentified-body cases, disaster victim identification and other investigations requiring the identification of human remains.
However, the distinction between facial reconstruction and positive identification remains critical. A facial representation generated by an AI model is an estimate, not proof of identity. Any potential match would require corroboration through appropriate investigative and forensic procedures, which may include DNA profiling, dental comparison, fingerprints, anthropological assessment and relevant ante-mortem records, depending on the circumstances of the case.
Research Findings and Further Development
The research has progressed through data collection, dataset preparation, model development, testing and evaluation. According to the report, the work has been published in peer-reviewed journals and presented at international conferences in 2026, including the International Conference on Pattern Recognition and the International Joint Conference on Neural Networks.
The researchers are continuing to improve the model's accuracy and assess its performance using larger and more diverse datasets, particularly those representing Indian populations.
This is an important consideration for forensic applications. Human cranial morphology and facial characteristics vary across populations, and a model trained on limited or insufficiently representative datasets may produce less reliable results when applied to other groups. Population-specific validation, transparent performance evaluation and careful assessment of error rates will therefore be important before the technology is adopted in operational forensic settings.
Implications for the Future of Forensic Science in India
The IIT Mandi initiative illustrates the growing intersection of artificial intelligence, forensic anthropology and computer vision. If further validated, such systems could provide investigators with an additional tool for generating leads in difficult identification cases.
Nevertheless, AI-generated results must be interpreted within the broader evidentiary framework. The reliability of the underlying data, the possibility of algorithmic bias, the quality of the input images and the consequences of incorrect matches will all require careful consideration.
The researchers' approach is intended to assist forensic professionals rather than replace expert judgement. Further testing on larger and more representative datasets will be essential to determine its practical effectiveness.
As forensic science increasingly incorporates computational methods, the challenge will be to ensure that technological innovation strengthens identification processes while maintaining scientific reliability, transparency and evidentiary integrity.