Friday, February 23, 2024
HomeAIAI development of "human detection" such as congestion avoidance and remote monitoring!

AI development of “human detection” such as congestion avoidance and remote monitoring!


Table of Contents

  • Implementation background
  • Overhead person dataset overview
  • Campaign overview

Implementation background

According to the() “Survey on the Global Surveillance Camera System Market” *1 released by Yano Research Institute in October 2021, after the corona crisis, in order to manage stores and facilities in a non-contact and non-face-to-face manner, More and more companies are using cloud cameras for remote monitoring.

In the global market for surveillance camera systems, the introduction of image analysis solutions using AI is expected to progress in each of the markets in Europe, the United States, Japan, and China. In addition, the demand for cameras for labor saving and unmanned operations is on the rise, and is expected to reach 6.4 trillion yen in 2026.

On the other hand, it is necessary to improve the accuracy of human detection for remote monitoring cameras that are used for customer analysis of crowds and visitors. In many cases, remote surveillance cameras are installed overhead so as not to interfere with the traffic of people, and to improve the accuracy of human detection, it is necessary to have image data that looks at people from a “bird’s-eye view”.

In anticipation of new machine learning needs for improving the accuracy of “face recognition” due to the pandemic, PIXTA has released a machine learning dataset of “images of Japanese people wearing masks” in 2021.

Furthermore, in response to the demand for non-contact and non-face-to-face remote monitoring and customer analysis through image analysis, which is progressing due to the corona crisis, we have started selling the “overhead person image” data set following the “image of Japanese people wearing masks”. .

*1 Source: Yano Research Institute Ltd., “Survey on the Global Surveillance Camera System Market (2021)” (announced on October 26, 2021)

Overhead person dataset overview

Dataset content

We sell two patterns of datasets, depending on the presence or absence of annotations *2 . The ready-made annotation has a bounding box attached to the whole body of the person, and the annotation data is provided in CSV format.

*2 Annotation: When image data is used as training data in machine learning and deep learning, the necessary information is attached to the image as a tag according to the content you want to develop.

Material condition

The 1000 bird’s-eye view image materials for sale are

  • Photos taken from a bird’s-eye view of a person
  • Japanese people
  • 1 to multiple people
  • Wide age range from children to seniors
  • Angle of view looking down from a high place (diagonal, directly above, etc.)
  • Centered on the whole body, including only the upper body


Bird’s-eye view person 1,000 points With annotation 165,000 yen (tax included)
Bird’s-eye view person 1,000 points Without annotation 99,000 yen (tax included)

Source: Example dataset with PRtimes annotation

Campaign overview

For companies and organizations that are considering purchasing photo data for machine learning on PIXTA, we will offer a free trial of 100 data sets from any of the machine learning datasets on sale on PIXTA for the first purchase. It will be held for a limited time.


Those who are considering purchasing materials for machine learning applications on PIXTA for the first time

Target Machine Learning DatasetDataset

  • Japanese people (with or without annotation)
  • Bird’s-eye view person (with or without annotation)


Please enter your comment!
Please enter your name here

Recent Posts

Most Popular

Recent Comments