Ever pondered the sheer number of individuals traversing the zebra crossing, particularly at the Shibuya scramble crossing? Delving further into the Shibuya Scramble Crossing, renowned as Shibuya Crossing, it emerges as a legendary pedestrian scramble intersection in Shibuya, Tokyo, Japan. Situated before the Hachikō exit of Shibuya Station, this vibrant crossing brings traffic to a standstill in all directions, resulting in a captivating spectacle as pedestrians inundate the entire intersection. There are countless number of pedestrians crossing after every interval of two minutes, enough to quickly fill up a football stadium. The phenomenon gave rise to its nickname “scramble,” as pedestrians cross from all directions.
What if we could analyze the number of people that are crossing after every set-interval using Object Detection Algorithm.The Object Detection algorithm enables us to identify the accumulation of individual who intend to cross. Examining the Shibuya scramble crossing under two conditions, during summer and on a rainy day. By feeding a time-lapse video (which contains almost 1700 frames) to the object detection model of the different situations. Surprisingly the number of people crossing in both situation are almost similar. Here’s an example of two different situations people crossing the Shibuya crossing.
The provided images represent individual frames extracted from two separate time-lapse videos. These frames capture the moment when people are crossing, revealing a similar number of individuals in both images. Consequently, we constructed a graph using the data from the two time-lapse videos to accurately depict the actual number of people crossing during each interval at the Shibuya crossing.
The images indicate a similar volume of people crossing in both situations. Additionally, we included data on the number of cars detected in each scenario. Interestingly, in the image depicting people crossing on a rainy day, there are more cars detected compared to the clear weather scenario. However, despite this difference, both graphs illustrate a significant number of people crossing the Shibuya crossing. It’s remarkable how this bustling pedestrian intersection stands out as one of Tokyo’s most iconic symbols worldwide.Due to its extensive advertising screens and bustling pedestrian activity, Shibuya Crossing is frequently likened to the Times Square intersection in New York and Piccadilly Circus in London. It is widely regarded as emblematic of Tokyo’s ultra-modern image projected globally.
Utilizing object detection algorithms, we can highlight the remarkable locations in Tokyo. However, when applied to time-lapse videos, these algorithms struggle to accurately detect the number of people or objects due to the rapid pace of the footage. Lower accuracy was observed initially. However, by reducing the speed of the time-lapse video, the accuracy of object detection improved. The algorithm achieves higher accuracy by analyzing each frame of the slowed-down video individually. Furthermore, providing each frame as separate images to the object detection algorithm can further enhance accuracy, allowing for the detection of a greater number of objects.
This blog post was created by Pooja Srinivasan and Parithosh Dharmapalan