Recently, an agent-based pedestrian simulation model was developed at the University of British Columbia to model detailed pedestrian interactions. The model was originally calibrated and validated with video data, collected at a signalized intersection in Vancouver. However, it is important to investigate the applicability of the model in different walking environments. The main objective of this study was to assess the model performance in handling pedestrian interactions at a scramble phase signalized intersection in Oakland, California. The intersection had four conventional crosswalks in addition to two diagonal crosswalks, used during the scramble pedestrian phase. Model parameters were calibrated with a genetic algorithm, which aimed to minimize the location and speed error between simulated and actual trajectories, extracted from the video sequence through computer vision. Validation results showed that the model was capable of reproducing pedestrian trajectories with high accuracy in regard to average location and speed error. The average location error for 271 pedestrians considered in the validation was 0.49 m, while the average speed error was 0.04 m/s. Detailed analysis of crossing speed for both conventional and diagonal crossings was presented, and the ability of the model to produce the same speed distributions observed in actual data was confirmed. Furthermore, the ability of the model to reproduce five interactions that were frequently observed in the data was assessed. Results showed that the model was capable of reproducing the actual behavior taken by pedestrians during these interactions with high accuracy, from 80% to 100%.