Classifying ethnicity Of the pedestrian using skin colour palette.

Abstract

Nowadays, with the emergence of big data, there is an increasing desire to analyze and understand them. In this area, the focus is on the challenge of identifying pedestrian attributes from CCTV images, especially in terms of ethnicity. Due to a lack of necessary characteristics, ethnicity classification is nigh impossible in this instance. Facial landmarks are a requirement for the existing approaches. Therefore, it is suggested to use the individual’s skin tones as features instead. Segmenting the skin area of each unique face adds multiple dominant colours to the colour palette, which are later employed as characteristics during classification. The P-DESTRE dataset, which provides pedestrian dataset and their properties, including their ethnicities, is used to demonstrate the viability of the suggested method. The accuracy percentage for distinguishing between Caucasian and Indian pedestrians using the P-DESTRE dataset and skin colour palette is 98%. The outcome demonstrates that ethnicity classification is possible when utilizing a colour palette as a feature. On this premise, it is still possible to identify a pedestrian’s ethnicity from CCTV footage even without the use of face landmarks.