Browsing by Author "Hirashkar, H."

Filter results by typing the first few letters
Now showing 1 - 2 of 2
  • No Thumbnail Available
    Item
    Matching Witness' Account with Mugshots for Forensic Applications
    (2018) Mohan, A.; Dhir, R.; Hirashkar, H.; Chittaragi, N.B.; Koolagudi, S.G.
    This paper proposes a system that can be used by the forensics department to identify and disclose criminal details automatically. The problem of matching the description of a suspect in a crime scene provided by an eye-witness to existing mugshots (mugshots represents photograph taken as someone is arrested) in the police departments criminal database is addressed in this work. Prominent features such as skin colour, size of nose lips, shape the size of eyes, and shape of the face are considered for discrimination of individual criminals. The witness fills in the description fields through which, most appropriate images are selected from an existing database. Images are scored on the basis of the degree of closeness to the given description, and most relevant images are displayed first followed by the rest. The classification of images based on explored facial features is done using extreme gradient boosting (XGBoost) supervised an ensemble learning method. Comparatively better performances are observed. � 2018 IEEE.
  • No Thumbnail Available
    Item
    Matching Witness' Account with Mugshots for Forensic Applications
    (Institute of Electrical and Electronics Engineers Inc., 2018) Mohan, A.; Dhir, R.; Hirashkar, H.; Chittaragi, N.B.; Koolagudi, S.G.
    This paper proposes a system that can be used by the forensics department to identify and disclose criminal details automatically. The problem of matching the description of a suspect in a crime scene provided by an eye-witness to existing mugshots (mugshots represents photograph taken as someone is arrested) in the police departments criminal database is addressed in this work. Prominent features such as skin colour, size of nose lips, shape the size of eyes, and shape of the face are considered for discrimination of individual criminals. The witness fills in the description fields through which, most appropriate images are selected from an existing database. Images are scored on the basis of the degree of closeness to the given description, and most relevant images are displayed first followed by the rest. The classification of images based on explored facial features is done using extreme gradient boosting (XGBoost) supervised an ensemble learning method. Comparatively better performances are observed. © 2018 IEEE.