<p>Addressing the heterogeneous traffic conditions, conflicting flow volume is an integral parameter in evaluating the traffic capacity of an unsignalized intersection. The present study evaluates legitimate manoeuvring patterns in heterogeneous traffic situations, considering spatial constraints and driving obstructions in conflict areas of unsignalized intersections in India’s tier-II cities. This enables accurate estimation of capacity at unsignalized intersections. The trajectory analysis of vehicular movements was conducted using the YOLOv5-DeepSort system to detect and track vehicles. The proportional spread of various manoeuvring patterns is obtained, assigning different coefficients to each, relying on conflict behaviour and traffic volume at the unsignalized intersections. Three models have been assessed: the simplified Indo-HCM capacity model, a modified capacity model designed for tier-II cities, and a regression-based capacity model. A total of 20 sites were identified in three tier-II cities, encompassing three-legged and four-legged intersections. The developed conflicting flow volumes indicate the actual traffic volume at conflict zones of unsignalized intersections in developing cities, thereby enhancing the efficacy of the regression capacity model during validation. The regression model, validated with R² values of 0.811 for three-legged and 0.915 for four-legged intersections, exemplifies better reliability and accuracy over the simplified Indo-HCM and modified capacity (tier-II city) models, which overestimate the traffic capacity at designated sites of the study. The proposed regression capacity model can be effectively implemented or utilised by traffic planners and designers to assess unsignalized intersection capacity in emerging and smart cities.</p>