Comprehensive Summary of 2D/3D Annotation Tools for Computer Vision

Annotation tools are the first step in processing raw data. Whether it is detection tasks, segmentation tasks, or 3D perception and point clouds, ground truth data must be created to supervise network learning. Enterprise-level annotation solutions are generally completed through internally developed tools or professional annotation teams, while an open-source and user-friendly annotation tool is crucial for individuals or small teams. The heart of autonomous driving has summarized commonly used annotation tools in the field, covering 2D detection segmentation, 3D detection segmentation, and multi-sensor calibration synchronization.

Detection Segmentation Calibration

1. Labelme

Project Address: https://github.com/wkentaro/labelme

Mainly supports:

• Polygon segmentation, semantic segmentation, 2D boxes, line annotations, point annotations (can be used for object detection, image segmentation tasks)
• Video annotation

Annotations are saved in JSON, VOC, and COCO formats;

2. LabelImg

Project Address: https://github.com/heartexlabs/labelImg

Three types of annotation files are available: PASCAL VOC, YOLO, CreateML, only supporting data annotation for object detection tasks.

3. CVAT

Project Address: https://github.com/openvinotoolkit/cvat

Free online interactive video and image segmentation annotation tool;

4. VOTT

Project Address: https://github.com/microsoft/VoTT

A JavaScript-based annotation tool for image object detection released by Microsoft, developed using React+Redux, and supports running on Windows and Linux platforms. The software also provides a method for automatic annotation using the faster-rcnn model trained with CNTK, which can significantly reduce the workload required for annotation.

5. EISeg

EISeg (Efficient Interactive Segmentation) is an efficient and intelligent interactive segmentation annotation software developed based on PaddlePaddle. It covers high-quality interactive segmentation models in various fields, including general, portrait, remote sensing, medical, and video. Additionally, annotations obtained from EISeg can be applied to other segmentation models provided by PaddleSeg for training, enabling the creation of high-precision models for customized scenarios, covering the entire process from data annotation to model training and prediction.

Old Version:

https://github.com/PaddleCV-SIG/EISeg

New Version: https://github.com/PaddlePaddle/PaddleSeg/tree/release/2.6/EISeg

Main features:

• Efficient semi-automatic annotation tool, already launched on multiple top annotation platforms
• Covers various vertical scenarios, including remote sensing, medical, video, 3D medical, etc.
• Multi-platform compatible, easy to use, supports multi-category label management

6. RITM

Samsung’s open-source interactive calibration tool RITM: https://github.com/saic-vul/ritm_interactive_segmentation

Used for click-based interactive segmentation, this model uses segmentation masks output from pre-trained inference models, which can not only segment a brand new object but can also start from an external mask and correct it;

Multi-Sensor Calibration

1. OpenCalib

Project Address: https://github.com/PJLab-ADG/SensorsCalibration

A toolkit open-sourced by SenseTime, supporting calibration between camera, lidar, imu, and radar:

The calibration tool supports checkerboard, circular, and square methods!

3D Detection Calibration

1. point-cloud-annotation-tool

Project Address: https://github.com/springzfx/point-cloud-annotation-tool

Mainly used for calibrating the box information of targets in 3D point clouds, supporting export in KITTI format and Apollo 3D format!

2. annotate

Project Address: https://github.com/Earthwings/annotate

Calibration based on the ROS framework, generating 3D detection boxes, xyz, and whl information;

3. 3D BAT

Project Address: https://github.com/walzimmer/3d-bat

Can be used to calibrate 3D targets such as cars, trucks, motorcycles, autonomous vehicles, and pedestrians, supporting annotation of any target with more than 10 points!

4. SUSTechPOINTS

Project Address: https://github.com/naurril/SUSTechPOINTS

Mainly supports:

• 9-degree-of-freedom box editing
• Object type/ID/property editing
• Interactive/automatic rectangular fitting
• Batch mode editing
• Perspective/projection view editing
• Multiple camera images, automatic camera switching
• Binary/pcd files of point cloud data
• Jpg/png image files
• Object/box/point coloring
• Focus mode, hide background, easily view details
• Stream play/pause
• Object ID generation

3D Point Cloud Segmentation Calibration

1. semantic-segmentation-editor

Project Address: https://github.com/Hitachi-Automotive-And-Industry-Lab/semantic-segmentation-editor

A web-based tool that supports image and point cloud formats for 2D and 3D data segmentation annotation!