What is Key Point Annotation? Its Features and Annotation Methods

1. What is the reason for the focus on keypoint annotation?

Key point annotation is mainly used for posture estimation and face recognition. In the annotation for posture estimation, points are inputted to estimate the position of joints such as shoulders, elbows, wrists, and knees. In order for AI to recognize posture changes, it is important for AI to learn the pixel information of images extracted from videos on a frame-by-frame basis and the coordinate information of each point, so that it can recognize changes in posture and facial expressions. Therefore, it is important to prepare key point annotation data using various images of postures and facial expressions. By doing so, the accuracy of AI in recognizing posture and facial expressions will improve. In addition, the more points that are inputted, the more complex changes can be recognized.

 

Next, we will introduce the reasons why keypoint annotation is gaining attention in pose estimation and face recognition.

1-1. Posture Estimation and Motion Capture:

Key Point Annotation is used to accurately capture body poses and movements. In posture estimation and motion capture, changes in nodal points such as joints are captured to digitize posture and body movements. In the past, in order to capture these movements, it was common to attach sensors to a person and use multiple cameras to capture their movements. This method requires a well-equipped environment to obtain data and involves measuring movements with multiple sensors attached. Due to these constraints, it was difficult to obtain data in narrow spaces, complex movements, and multiple target movements.

 

With the emergence of AI, it has become possible to obtain keypoint annotations for this data, making it possible to estimate movements from two-dimensional data such as images and videos. This allows for posture estimation and motion capture without relying on location or measurement equipment. With the ability to achieve high accuracy with just one camera, the applications for motion detection have expanded to various situations.

 

 

1-2. Face Recognition:

By inputting points at specific locations on the face, AI can learn changes in facial expressions and differences in individual face shapes. By setting points in detail, such as at the corners of the eyes, mouth, cheeks, and chin, more accurate recognition becomes possible. By utilizing keypoint annotation, it is possible to capture detailed and complex features of the subject that may be difficult with other annotation methods. Even if bounding box annotation is used to surround the entire face and recognize it with AI, it may be difficult for the AI to recognize the same person if they are wearing glasses, a mask, or a hat. In such cases, if the keypoint annotation can capture the position of the corners of the eyes, nose, chin, and ears, which are unique to that person, accurate recognition of even just a part of the face becomes possible.

 

 

2. How to do Key Point Annotation

At Key Point Annotation Project, we will proceed with the work through the following steps.

2-1. Data Collection:

Collect the dataset to be used for keypoint annotation. Gather data such as images and videos that contain the subject or movement you want to detect. Additionally, for human pose estimation, there are also methods to utilize existing datasets such as the COCO dataset.

2-2. Selecting a Data Annotation Tool:

Select a tool suitable for keypoint annotation. There are various tools available, such as open source and commercial tools, so choose a tool based on various factors such as usability and the purpose and situation. Depending on the size of the project, it is also recommended to choose a tool with project management capabilities.

2-3. Setting Points:

Decide on the points to input (annotation specifications) for the subject you want to learn. Depending on the subject, determine how to mark the points to obtain the best data. Also consider creating documents such as definitions and guidelines to ensure consistency in the annotation process. If possible, it would be even better to perform a small amount of annotation with actual data before starting the work and test it with AI.

2-4. Implementation of Data Annotation:

Use the annotation tool to annotate the set points on the image. It is important to accurately annotate each point. However, pursuing too much accuracy can take a lot of time and effort to prepare a large amount of data. Therefore, it is a good idea to determine the acceptable range of error beforehand. In addition, as the work progresses, there are often cases where it is difficult to make judgments or where things do not fit into guidelines or definitions. In order to ensure the quality, it is important to assign a responsible person and make efforts to update the guidelines.

2-5. Data Verification and Correction:

Check the annotated data and make necessary corrections. It is important to establish a system to ensure quality, such as setting the frequency of checks based on quality requirements and assigning a dedicated QA checker if necessary.

3. Summary

As we have seen so far, keypoint annotation is applied to various fields of image recognition, especially pose estimation and face recognition. As for its application to business, it can be used for detecting abnormal movements of drivers in the transportation industry, as well as analyzing human movements in sports, medical, and educational fields. By utilizing AI trained with keypoint annotation data, it is possible to solve various challenges that were difficult to address with conventional technology.

 

In order to ensure the accuracy of AI recognition, it is essential to create accurate annotations and high-quality datasets. If it is difficult to create keypoint annotations in-house, utilizing an external vendor can also be effective.

4. Human Science's Data Annotation Outsourcing Service

4.8 million records of teacher data creation

At Human Science, we are involved in AI development projects in various industries such as natural language processing, medical support, automotive, IT, manufacturing, and construction. Through direct transactions with many companies including GAFAM, we have provided over 48 million high-quality training data. We handle various annotation projects regardless of industry, from small-scale projects to large-scale projects with 150 annotators. If your company is interested in introducing AI but unsure where to start, please consult with us.

Resource Management without Using Crowdsourcing

At Human Science, we do not use crowdsourcing and instead directly contract with workers to manage projects. We carefully assess each member's practical experience and evaluations from previous projects to form a team that can perform to the best of their abilities.

Utilize the latest data annotation tools

One of the annotation tools introduced by Human Science, AnnoFab, allows customers to check progress and provide feedback on the cloud even during project execution. By not allowing work data to be saved on local machines, we also consider security.

Equipped with a security room within the company

At Human Science, we have a security room that meets the ISMS standards in our Shinjuku office. This allows us to provide on-site support for highly confidential projects and ensure security. We consider confidentiality to be extremely important for all projects at our company. We continuously provide security education to our staff and pay close attention to the handling of information and data, even for remote projects.