What is annotation

アノテーションとは、画像、テキスト、音声などのデータに対して、関連する情報を注釈としてタグ付けする作業のことです。AI開発における教師あり学習では、教師データ(正解データ)が必要になります。精度の高いAIモデルを開発するためにデータへの正しいタグ付けが非常に重要です。正確にタグ付けできていないデータを取り込んでも、AIは正しく学習できません。正しいタグが付いたデータを学習することで、正しくパターンを認識できるようになります。そのため、データの量だけでなく、データの質(アノテーションの精度)も非常に重要な要素となります。アノテーション作業は、膨大な数の画像一枚一枚に対して行うため、AI開発サイクルの中で最も多くの工数を要します。

Service overview

We provide high-quality, low-cost, and short delivery time annotation service that associates the "position" of the object to be recognized in the image with its "label name". It supports not only 2D annotation (Bounding Box, Semantic Segmentation) for still images obtained from various cameras, but also 3D annotation for 3D point cloud data obtained from LiDAR. We also flexibly respond to special requests, such as using the annotation tool specified by the customer.

Bounding Box

Enclose the object in the image with a rectangle (bounding Box) and tag it with a class that indicates what the object is and the coordinate information of the rectangle. Class classification and output format (json format) can be freely customized.
Class examples: 28 classes such as cars, trucks, buses, people, motorcycles, bicycles, traffic lights, road signs, etc.

Semantic Segmentation

All pixels in the image are tagged with which class they belong to. Class classification, color specification for each class, output format (json format), etc. can be freely customized.
Class example: 37 classes such as cars, people, bicycles, traffic lights, road signs, white lines, roads, sidewalks, buildings, sky, etc.

3D Point Cloud Annotation

Enclose the object in the 3D point cloud data (3D point cloud) obtained from LiDAR with a rectangular parallelepiped, and tag the class indicating what the object is and the coordinate information of the rectangular parallelepiped.

Business process

Document download

Please download the materials related to the annotation service from here. Detailed information such as service contents and processes is provided.

Inquiry

If you have any questions or concerns about the annotation service, please feel free to contact us here.