A Comparative Study of Deep Learning Models for Photo-to-Sketch Recognition with CAT Preprocessing

View at publisher Summarise with VIRAT
Authors: Philomina Princiya Mascarenhas, M S Sannidhan, Aashna Mathias, Dabis Camero, Jason Elroy Martis
Cited by: 0
References: 19

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

In recent years, facial recognition systems have become widely used in many applications, especially in biometrics and law enforcement. Advances in the field have led to the development of pre-trained deep learning models that achieve good accuracy even when training data is limited. Earlier studies have shown that these models perform well on original photographs but struggle when the input comes from other domains such as sketches. In crime investigations, officers often have only the suspect's sketch, and using pre-trained models directly on such sketches gives low accuracy because important texture and lighting details are missing. To address this issue, our study applies a Contour Accentuation Technique (CAT) to enhance the texture information in sketches. This improves the clarity of sketch features and supports better learning by deep learning models. To evaluate the effectiveness of CAT, we conduct a comparative study using multiple recognition models with and without CAT. The models are tested under a unified setup with cross-validation. The results show that the use of CAT improves face recognition performance, with an average accuracy gain of 7.7%. The working model of our paper is found in https://github.com/Princiya1990/faceRecognition.

Fields of Study & Subjects

Face recognition and analysis Face and Expression Recognition Biometric Identification and Security

Publication details

Journal / source
International Conference on Communication, Computing and Electronics Systems (ICCCES)
Pages
2186–2191
Publication date
21 January 2026
Language
EN