The 8 Best Point Cloud Labeling Tools for 2024

February 23rd, 2024 - 6 min -

It’s simply not enough to give a computer a large amount of data and expect it to learn – data has to undergo preparation for computers to find patterns within it. That’s where data labeling platforms come in. For many ML models, it is enough to rely on images. However, automotive and robotics companies increasingly use a combination of lidar, radar, and camera sensors. The more sensors they use, the more accurate and safer their ML models become.

Find out which point cloud and multi-sensor annotation tools are popular in 2024. We compare their features and prices to help you choose the best tool for your data annotation needs.

Here are the platforms we’ll cover

  • 1
  • 2
  • 3
  • 4

    SageMaker Ground Truth

  • 5
  • 6
  • 7
  • 8
  • 9
  • 10
    Telus International
  • 11
  • 12
  • 13

Yes, this is us. So, we won’t pretend to be unbiased here and promote ourselves to the top of this list. Give us a spin for yourself with our free trial instead, no strings attached. launched as a data labeling solution designed explicitly for segmentation. We had early success with customers in the robotics and autonomous driving industries. This prompted us to address the complex needs of autonomous navigation, leading us to introduce point cloud support.

Through close collaboration with our customers, we saw that manually labeling multiple modalities is time-consuming, costly, and ineffective. Therefore, we shifted our focus from being the leader in segmentation to providing the best multi-sensor labeling capabilities.

Combining labeling various sensors, including lidar and images, into a single interface leads to improved efficiencies and consistency in a dataset. This approach allows for consistent track IDs across time and modalities, enables labeling multiple sensors with just one click, and facilitates the fusion of information from various sensors.

Overlay images with point clouds for better data labeling supports:

  • Image labeling (incl. sequences)
  • Point cloud labeling (incl. sequences)
  • Multi-sensor labeling in a single interface

Key features:

  • Single interface to label multiple sensors
  • 3D-to-2D projection
  • Zero-shot models
  • Consistent track IDs across sensors
  • Upload your own models with model-assisted labeling
  • No size limit on point clouds
  • Merged point cloud for higher accuracy
  • Batch mode to label dynamic objects quickly
  • Real-time synced camera images or image underlay for faster object recognition


  • Free for academic users
  • Starts at 9.600/year
  • Pricing is based on hours of active labeling, with unlimited users, data, and labels


Deepen, founded in 2017 by Mohammad Musa and Andrew Lee, is a data labeling service and tool for autonomous systems in machine learning. The company emphasizes safety and offers both 2D and 3D annotation capabilities.

Key features in the image labeling platform include the standard ML-assisted labeling options like superpixel, autosegment, and SAM for image labeling, alongside automated pre-labeling for 80 standard classes.

In the point cloud interface, Deepen features fused cloud and batch mode for point cloud labeling, but compared to its 2D capabilities, it lacks automated labeling functions for point clouds. The platform supports point clouds with up to 500 million points. The multi-sensor support is limited to an image-based modal alongside the point cloud labeling.

Deepen stands out with its calibration tools and involvement in the Safety Pool, a safety initiative in partnership with the University of Warwick. Additionally, the Deepen Exchange platform allows users to trade or share datasets.

Support and Key Features:

  • Image and point cloud labeling
  • Pre-labeling images for 80 standard classes
  • Point clouds up to 500 million points
  • Sensor calibration tools


  • Custom pricing only
  • Calibration starts at $500/month


Kognic, initially Annotell, originated in Sweden in 2018 and was founded by Daniel Langkilde and Oscar Petersson, with its headquarters still in Sweden. The platform aids product teams in developing and overseeing custom data pipelines, offering functionalities for data exploration, organization, and interpretation.

A standout feature is the data curation overview, enabling the selection or grouping of data before dispatching to labelers. Kognic enhances data value for model training through various pre-annotation methods, followed by a human-in-the-loop workflow.

Kognic’s multi-sensor capabilities are limited to a synchronized image viewer in the point cloud interface. Viewing 2D and 3D data at the same time helps to give more context to labelers.

The platform’s QA section includes an insights tab that presents metrics about the data, labeling team, and test outcomes.

Support and Key Features:

  • Point cloud labeling
  • Dataset curation tools
  • Support for pre-label methods


  • Custom pricing only

SageMaker Ground Truth

Amazon introduced SageMaker Ground Truth in 2017. This tool supports annotating images, text, and point clouds for various applications.

A unique feature in SageMaker is the assistive labeling tool, Snapping. It enables workers to encapsulate an object within a cuboid, which Ground Truth’s autofit tool adjusts to fit around it. This functionality also allows you to align cuboids with surfaces like the ground or sidewalks.

Like, SageMaker allows for the overlay of images with point clouds, facilitating a sensor fusion view. This integrated approach lets workers refine annotations in both 3D scenes and 2D images, with real-time projection of adjustments across views, enhancing accuracy and context in annotations.

Support and Key Features:

  • Image, text, and point cloud annotation support
  • Sensor fusion for overlaying 2D and 3D data
  • Autofit and ground-snapping


  • Freemium model
  • A (very) long list of custom options


Founded in 2017, CVAT stands out as an open-source solution with a lean team of 10 core members, offering labeling tools for images and videos. Users can utilize CVAT’s hosted service or opt for a self-hosted setup, providing image and video annotation functionalities.

For image labeling, CVAT offers various shapes for object labeling, such as bounding boxes, vectors, and skeletons for tracking human positions. Point cloud annotation is limited to 3D bounding boxes.

CVAT enhances efficiency for image labeling with assistive tools like autosegment and SAM. They also integrate popular models for automated image labeling, such as YOLO and a text or face detection model. The platform supports model-assisted labeling, allowing users to upload their custom labels. Despite these features for image labeling, point cloud labeling remains constrained, with 3D bounding boxes as the only option and limited interpolation tracking for dynamic objects in 3D scenes.

The platform includes a modal to visualize synchronized images alongside 3D scenes but lacks real-time label synchronization between 2D and 3D views.

Support and Key Features:

  • Image and skeleton labeling,
  • Point cloud labeling is restricted to cuboids
  • Pre-installed models for automated annotations in images
  • Options for cloud-based or self-hosted deployment


  • Free plan for up to 3 users
  • Paying plans start at $33 per month per seat
  • Custom pricing for self-hosted solutions


Founded in 2017, Dataloop specializes in managing the complete generative AI cycle. They do still offer the annotation platform for images and point clouds.

The platform’s point cloud labeling capabilities are limited to 3D bounding boxes and polylines. It includes a 2D image viewport in the 3D point cloud interface. Like Amazon, Dataloop features ground-snapping for bounding boxes.

A unique offering from Dataloop is the “Point Cloud Focus” feature, which allows users to zoom into specific areas of interest in a point cloud and hide irrelevant points. This helps reduce noise and concentrate on a single area or object within a scene.

Dataloop also supports model-assisted labeling, enabling the running of models on datasets to generate annotation predictions. These pre-labeled items are then reviewed by annotators, greatly enhancing productivity.

Support and Key Features:

  • Image and point cloud labeling with bounding boxes and polylines
  • Model-assisted labeling
  • Ground-snapping
  • Point cloud focus


  • Custom pricing


Supervisely adopts a unique strategy by proposing an ecosystem for building a customizable platform through custom or pre-built apps rather than a standard data labeling product. This approach is designed to provide a more flexible and tailored solution for image, video, 3D, and medical imaging annotation.

The platform’s point cloud interface is equipped with autosegment features and provides an image viewport overlayed with point clouds and syncronized bounding boxes, handling point clouds with up to 1 million points.

The app ecosystem for point clouds primarily consists of data import and export applications, including support for exporting point clouds to the KITTI format. However, the only app available for point cloud labeling is the main labeling interface.

Support and Key Features:

  • Support for images, point clouds, and DICOM files
  • A library of apps for various use cases
  • Point clouds up to 1 million points


Supervisely’s pricing model is based on the number of modules and users, with point cloud labeling features being limited in the initial two tiers and fully accessible only under the enterprise plan.

  • Free plan for up to 2 members
  • $199/month starting at 3 members
  • Custom pricing for the Enterprise plan


Scale, established in 2016 and securing over $600 million in funding, stands out in the data labeling industry. While it offers a self-serve option, Scale primarily caters to enterprise clients seeking hands-off data labeling solutions.

Scale Studio, the self-serve platform, enables labeling across images, text, videos, PDFs, and audio on a pay-as-you-go basis. It features ML-assisted tools, including a function similar to superpixel and autosegmentation.

However, point cloud labeling is exclusive to enterprise plans.

Clients can work with Scale’s internal team or an external labeling workforce for their data labeling needs.

Support and Key Features:

  • Support for images, text, video, PDFs, and audio
  • Data curation tools
  • Pay-as-you-go pricing model


  • Self-serve model with pay-as-you-go pricing based on uploaded images and labeled units
  • Enterprise plan (including point cloud labeling) with custom pricing

Worth mentioning was founded in 2017 and was acquired 2 years later by dSpace in 2019. Under the umbrella of the dSPACE group companies, will invest in the core tasks of artificial intelligence (AI) applications and cloud-based tools. It will further develop its existing products as an integral part of the dSPACE product range. It still offers its labeling services to other ADAS and AD projects.

Telus International, Sama & Appen

Telus International, Sama & Appen have shifted their focus from offering a standalone labeling platform to only providing end-to-end services. These services include data collection, annotation, and validation, delivering a full suite of solutions to meet various data labeling needs without the option to license just the platform for independent use or with diverse labeling teams.

The open-source platform Scalable was discontinued at the end of 2023. It was a versatile annotation platform supporting 2D and 3D data labeling. BDD100K was labeled with this platform.

Share this article