dlabel.org

What It Is

dLabel is a data annotation service and platform for AI and computer vision projects, positioning itself around “training data prepared for AI by annotation experts.” According to its website, it has offices in Canberra, Australia and Tallinn, Estonia, with customers including CSIRO, AIS, and research teams worldwide. It is more of a combined “service + tooling platform” than a purely self-serve annotation tool.

Core Capabilities

dLabel covers mainstream visual annotation tasks. For images, it supports bounding boxes, polygons, semantic segmentation, keypoints, and attribute labeling. For video, it supports frame-by-frame annotation, keyframe interpolation, and object tracking across video sequences. The platform showcases capabilities such as semantic segmentation, spline annotation, pose estimation, smart segmentation, video tracking, and quality assurance, and supports common computer vision data formats including COCO, YOLO, and Pascal VOC. Its AI features are mainly reflected in AI-Assisted Labeling, including smart segmentation, semi-automated tools, and automatic segmentation to improve annotation efficiency; however, the website does not disclose the specific models, algorithms, or automation accuracy.

Pricing and Service Model

In terms of pricing, dLabel does not publish plans, unit prices, free quotas, or a trial entry point. Its process is for users to submit their dataset and requirements and then receive a detailed quote, with the website stating that responses are typically provided within 24 hours. As such, it is better suited to project-based procurement with a clear budget and more complex requirements, rather than small teams that want to sign up and start immediately on a self-serve basis.

Pros and Cons

Its strengths lie in end-to-end coverage from raw data to production-ready annotated datasets, with a broad range of task types. It is particularly suitable for complex computer vision tasks such as image segmentation, video tracking, and pose estimation. Built-in QA and export to common formats are also helpful for downstream model training. The limitation is that publicly available information is relatively sparse: it does not clearly explain data privacy and compliance measures, API/SDK integration, QA sampling ratios, delivery timelines, or quantifiable quality metrics, and there is no information on Chinese-language support.

Best Fit and Access from China

dLabel is suitable for computer vision R&D teams, research institutions, and projects in areas such as autonomous driving/traffic analytics, industrial inspection, remote sensing, or medical imaging that require professional human annotation and quality control. The website does not disclose details about access from China, network stability, or payment methods. Before formal procurement, it is advisable to confirm access, contracts, invoicing, cross-border data transfer, and confidentiality terms. If localized communication is required or the data should not leave China, domestic data annotation providers may be worth comparing; if you prefer a self-serve platform, alternatives such as Label Studio, CVAT, Labelbox, and SuperAnnotate can be evaluated.