Comparison with Alternatives

Supervisely vs competitors

Learn how Supervisely is different from other products on the market and find if the platform is the best fit for you.

Trusted by Fortune 500. Used by 65,000+ companies and researchers worldwide

A detailed feature comparison

You probably end up here because you are searching for a labeling platform. There are so many options to choose, from developing an in-house solution to open-source projects and enterprise-grade products.

To make your job easier, we have put together a detailed overview of how Supervisely is different from other solutions.

Alternative solutions
User friendly web-based GUI
User Interface is something that our customers often like
✅  Yes
✴️  Rarely
Interface customization
Customization (including Interface customization) is a strong side of Supervisely
✅  Yes
✴️  Rarely
Extensibility of features via Supervisely Apps
Supervisely is a platform, Apps provide us (and our users) easy way to add features and new functionality
✅  Yes
🚫  No
App development on request
App Development on a customer request is a standard practice for us. Moreover, in the majority of cases, the development is free of charge (the reason is that both sides benefit - (1) customer gets an App (2) Supervisely gets a new App in Ecosystem)
✅  Yes
🚫  No
Ecosystem of Apps: hundreds of apps that extend core functionality
Apps in Ecosystem address various tasks related to Computer Vision such as Quality Assurance of Data, Performance Analysis, AI assisted labeling, Neural Network training / inference and more
✅  Yes
🚫  No
Growth of features over time
Today all competitive solutions are essentially products. As a code base of any product grow, it's harder to add more features. On the other hand, Supervisely is a platform. Features are implemented as Apps. Apps are easy to create. Moreover, Apps may be created by Supervisely Team, Partners or Customers
✅  Exponential
🚫  Logarithmic
Price depends on the modules and number of concurrent users (not number of labeled objects)
When (1) the price depends on the number of objects labeled and (2) annotation at scale is taken place - the annotation software may cost millions of dollars per year - it's not the case with Supervisely
✅  Yes
✴️  Depends on usage
Customer Success Team
A lot of companies have such teams. However, due to the fact that Supervisely is a platform, it's way easier for us to implement missing features as Apps - to make sure that Supervisely actually solves the needs of Customers
✅  Yes
✴️  Rarely
Free evaluation period
There are several ways to check if Supervisely is a good fit for a Customer. For quick tests Community Version of Supervisely (supervise.ly) may be used - just sign up and start testing. However, we highly recommend participating in the Free Evaluation of Enterprise Version - in this case we create Dedicated Slack Workspace for direct communication, provide installation instructions for on-premise (self-hosted) installation and license key. By default, an Evaluation Period is 1 month long (though it may be easily extended if more testing time is needed). Enterprise Version has more features and is free of some limitations of Community Version. Most importantly, Supervisely Dev Team helps Customers with any tasks related to Supervisely - it may be a help with installation, configuration, customization (for example, if a certain feature is missing, Supervisely Team will create an App with functionality needed), how to use questions.
✅  Yes
✅  Usually

Image Labeling

Spatial labeling
Rectangles, polygons, masks, points, keypoints, brush
✅  Yes
✅  Yes
Categorical labeling for images & objects (with tags)
✅  Yes
✅  Yes
Custom text description for images & objects
✅  Yes
✅  Yes
High resolution images
In Supervisely there is a specialized Labeling Interface to efficiently label and navigate over high resolution images
✅  Yes
✴️  Rarely
Interactive segmentation (Smart Tool)
Smart Tool is a way to label masks in semi-automatic way (10x speedup in some cases)
✅  Yes
✴️  Rarely
Smart Tool training
Smart Tool is essentially a neural network that can be adopted (trained) for specific domain (object)
✅  Yes
🚫  No
Filtering of images & objects
✅  Yes
✴️  Rarely

Video, LiDAR and DICOM Labeling

Working with video directly (without splitting into frames)
Most commercial tools split videos into images first and then allow to label these frames (images). This approach is hard to scale. Supervisely allows to directly label raw videos. Though, splitting into frames option is also available
✅  Yes
✴️  Rarely
Spatial Labeling (rectangles, polygons, masks, points)
✅  Yes
✴️  Rarely
Categorical labeling for frames & objects (with tags)
✅  Yes
✴️  Rarely
Semi-automatic Object Tracking (State-Of-The-Art Models)
One frame is labelled and then bounding boxes automatically propogated to the subsequent frames. In Supervisely we rely on State-Of-The-Art class-agnostic tracking algorithms that usually ensure high accuracy
✅  Yes
✴️  Rarely
Interactive Segmentation (Smart Tool)
Allows to efficiently label objects with masks on multiple frames
✅  Yes
🚫  No
3D spatial labeling (cuboids, points, segmentation), Categorical Labeling (tags on objects), Multiple photo-contexts & sensor fusion, Point Cloud episodes (sequence of frames)
Current situation is the following. To organize annotation at scale for self-driving related industries a company may consider two option (1) free open source (not scalable) (2) solutions from scale.ai or amazon (since their prices depends on the # of labeled objects - cost may easily be in millions of USD per year). Under these conditions, Supervisely becomes promissing alternative
✅  Yes
✴️  Rarely
Working with DICOM directly (without splitting into frames)
There are a few powerful open source solutions available (for example, MITK), the main problem with them is scalability - those are desktop solutions. Supervisely is probably the only web-based solution on the market that supports working directly with DICOM and provides comparable set of features
✅  Yes
✴️  Rarely

Data organization & management

Let’s describe the current status on the market. Open source annotation tools don’t pay too much attention to Data Organization / Management or to collaboration at scale. Commercial products, on the other hand, have to deal with it. However, Commercial products address Data Organization / Management in a very “narrow sense” - they usually assume that “life ends” after the data is labeled and exported. This assumption is usually false, because we may want to utilize labeled data later. In other words, there should be a way to efficiently manage, query, transform raw and labeled data - for that, Data Organization and advanced Data Management tools are needed.
Basic data management
includes basic operations on workspaces/projects/datasets like copying, moving, sharing, removing, backing up
✅  Yes
✴️  Sometimes
Advanced data organization
Teams, Workspaces, Projects, Datasets...
✅  Yes
✴️  Rarely
Import from popular formats
Supervisely, Pascal VOC, COCO, Kitti, ...
✅  Yes
✴️  Rarely
Export to popular formats
Supervisely, Pascal VOC, COCO, Kitti, ...
✅  Yes
✴️  Rarely
Advanced data management
rich set of operations on workspaces/projects/datasets and corresponding images / videos / point clouds (example, Data Commander)
✅  Yes
🚫  Almost never
Transformations of raw and annotated data
sliding window, transformations of images (cropping, rotations,...), video to images, images to videos, transformations / merging of objects classes
✅  Yes
🚫  Almost never
Data Querying - of sorts of filtering, sampling
Apps, DTL, Data Commander
✅  Yes
🚫  Almost never

Collaboration at scale and Quality Assurance

User roles & permissions
✅  Yes
✅  Usually
Task management system for Labelers (Labeling Jobs)
In addition to GUI, for automation and customization of Labeling at scale processes, Supervisely SDK may be used.
✅  Yes
✴️  Sometimes
Review of annotation tasks
✅  Yes
✴️  Sometimes
Feedback for annotators (issues)
✅  Yes
✴️  Sometimes
Exams to test Labelers
✅  Yes
✴️  Rarely
Consensus labeling
Consensus labeling is task dependent - there are many ways to calculate consensus and to merge annotations from different Labelers. In other words, there is no “single best solution for consensus labeling”. Since Supervisely relies on Apps to implement “custom logic”, it is relatively easy to create an App that implements required “consensus logic”. Having said that, it’s important to note that “task dependency of consensus labeling” occurs in the final stage - once Labers performed the annotation. In Supervisely to set up a process when different Labelers annotate the same images is easy and straightforward (via Labeling Jobs). In addition, there are Apps that actually help with merging Annotations from different Labelers. Furthermore, if consensus is calculated by “given algorithm” then Supervisely SDK may be utilized to automate the process
✅  Yes
✴️  Rarely
Labelers statistics
✅  Yes
✴️  Rarely
Statistics related to data & labels
In Supervisely, we have relatively rich statistical tools to monitor (1) Data statistics - # of labeled images, # of objects (across classes and overall), # of object instances (2) Labelers performance statistics. However, it’s important to understand that really useful statistics may be task dependent - tools to perform deep analysis of labeled data for object detection task may be very different from tools to perform analysis of data for classification task (especially if the number of categories is huge). Supervisely Apps help here a lot - there are dozens of Apps in Ecosystem already that provide various ways to visualize and summarize data for various tasks.
✅  Yes
✴️  Rarely
Interactive statistics & data exploration
Statistics should be closely connected to the data. Another way to say it, is that statistical tools should be interactive. Most Apps in Supervisely Ecosystem that provide statistical analysis are interactive - while user inspect statistical properties of data, he can see the actual images that exhibit certain statistical properties.
✅  Yes
✴️  Rarely

Human-in-the-loop (HIL)

We use the following definition of Human-In-the-Loop (HIL) here: HIL is a process of obtaining initial predictions in automatic way and providing Labelers with the predictions so that the Labelers mostly perform corrections of annotations rather than annotating from scratch
HIL via SDK/API or GUI - Importing predictions from customer model and using the predictions as a "starting point" for the Labelers
This is the simplest form of human-in-the-loop approach. Any commercial software should support it
✅  Yes
✅  Yes
HIL via Models integrated in Supervisely - initial predictions for Labelers may be obtained either by (1) using pre-trained models available in Supervisely or by (2) training model in Supervisely and running inference
The fact that training and serving of Neural Networks is supported in Supervisely greatly simplifies implementations of various Human-In-the-Loop scenarios. Although, Data Labelling and Model Training functionality are tightly related, alternative solutions focus either on Labelling Tools or Model Training Tool, but not on both
✅  Yes
🚫  No
HIL via integration of Customer' model / model architecture
If there is a need to integrate custom model or custom model architecture so that other users of organisation could serve or train it during Human-In-the-Loop process, Supervisely provides guides to perform such integrations
✅  Yes
🚫  No

Automation & Customization

Rest API (HTTP API)
Language independent way to interact programmatically with Supervisely
✅  Yes
✅  Usually
Python SDK
Python SDK is a wrapper around REST API. All Apps in Supervisely Ecosystem relies on Python SDK to interact with Supervisely. Python SDK (as well as REST API) is a fundamental building block to (1) perform any automation (2) make it easy to develop Supervisely Apps
✅  Yes
✅  Usually
App Development
Supervisely Apps extend the core functionality of Supervisely (In general, the Value that Supervisely Apps provide to the end-user is comparable to the Value that Applications bring to an Operating Systems). From a user perspective - an App is indistinguishable from User Interface of Supervisely - that is when it comes to User Interface Apps are very powerful. Whenever there is a need to create a Custom Labeling Interface (or extend the existing one), organize Custom Workflow, visualize task dependent Statistic or integrate new Neural Network Architecture - Apps are the way to go. In that regard, Supervisely might be considered as Operating System and Apps that is something that make the Operation System useful to the end user. Due to the importance of Apps, Supervisely provides guides and tools for Enterprise Customers to create their own private Apps.
✅  Yes
🚫  No

Neural Networks

Neural Networks are just github repositories (that implement State-Of-The-Art Models) integrated in Supervisely as Apps. What it essentially means is that any SOTA model available on github is a potential Supervisely App (to transform a repo into an App a little bit of integration code is needed). This approach ensures that the latest SOTA models are available in Supervisely Ecosystem
Leveraging customers' machines with GPU for model training & serving (Supervisely Agent)
Supervisely Agent is an open source tool that (among other things) enables to connect users’ machines with GPU to Supervisely. This connection allows users to initiate GPU related computation (Training and Inference of models) from Supervisely (from web browser) and leverage connected user machines to perform these computations. In other words, users can train, serve, deploy models right from the web browser. The beauty here is that a user machine with a GPU may be a machine at home, at office or machine in the cloud.
✅  Yes
🚫  No
Clasification (State-Of-The-Art model architectures) - Training & Serving
Alternative solutions focus either on Labelling Tools or Model Training Tools, but not on both. Supervisely Team believes that Labelling Tools and Model Training Tools are closely connected and should be available under single platform
✅  Yes
🚫  No
Detection (State-Of-The-Art model architectures) - Training & Serving
The same as above
✅  Yes
🚫  No
Semantic Segmentation (State-Of-The-Art model architectures) - Training & Serving
The same as above
✅  Yes
🚫  No
Instance Segmentation (State-Of-The-Art model architectures) - Training & Serving
The same as above
✅  Yes
🚫  No
Detection Models on 3D Point Clouds - inference
✅  Yes
✴️  Sometimes
Interactive Segmentation Models (Smart Tool) - Inference & Training
✅  Yes
✴️  Sometimes
Class-agnostic Object Tracking Models - Inference from Video Labeling Interface
✅  Yes
✴️  Sometimes
Model Training Dashboards
Models architectures integrated in Supervisely come with Interactive Dashboards to represent Training dynamics in a visual user-friendly way
✅  Yes
🚫  No
Various User Interfaces for Model Inference
Once a Model is integrated in Supervisely, it can be leveraged via various Graphic User Interfaces. Examples (1) Smart Tool may be used either in Image Labelling Interface or in Video Labelling Interface - the same model is an “engine” behind very different Labelling Interfaces (2) Neural Networks integrated via Apps - on option to run apply model is to run Inference App on the Project level, another option is to apply model inside Labelling Interface to process individual images
✅  Yes
🚫  No
Performance Analysis (example, "Object detection metrics" App)
Tools to conduct an analysis of model performance in an interactive and systematic way (Interactive Confusion Matrix, Edge case analysis and more)
✅  Yes
🚫  No

Synthetic Data & Vertical Solutions

Companies that are involved in annotation business have a very interesting choice, i.e. they may choose to be (1) horizontal - their solutions have to be general and work good for most industries (2) vertical - their solutions are adopted for specific domain / industry / task (for example, Trax specializes on Retail). In Supervisely we don’t have to choose since Supervisely is a platform (horizontal solution) but certain App Collections adopt Supervisely to various industries (vertical solutions). Over time, Supervisely will support more and more verticals (industries)
Synthesis of images ("Flying Object" App)
✅  Yes
🚫  No
Synthesis of videos ("Synthetic videos for tracking" App)
✅  Yes
🚫  No
Action Recognition Labeling
Collection of Apps that make it convenient to organize labeling at scale of Action Recognition on videos. This vertical solution (1) reduces complexity of labeling (since only relevant annotation instruments are shown to the user) (2) contains tools to perform Quality Assurance specifically for Action Recognition task. As for competition, open source tools do help with Action Recognition, but collaboration, scalability and quality assurance will be a problem. Commercial products rarely (none we are aware of) provide specialized solutions for this task.
✅  Yes
🚫  No
Product Labeling for Retail
Collection of Apps that leverages Metric Learning to automate categorisation (association of product identifier) of products on Supermarket shelves. This Collection of Apps nicely illustrates how powerful Online Learning can be. As was mentioned before, there are competitive solutions on the market for this industry
✅  Yes
✅  Yes

Infrastructure

On-Premise (self-hosted) installation
✅  Yes
✅  Usually
External Authorization - OpenID, LDAP
✅  Yes
✴️  Sometimes
Remote Storage - Azure Blob Storage, Google Cloud or any S3 compatible storage (i.e. AWS S3)
✅  Yes
✴️  Rarely
CDN
Supervisely allows to use CDN services to fetch images and videos in labeling tools to speedup data fetching from remote locations
✅  Yes
✴️  Rarely

SUPERVISELY FOR ENTERPRISES

On-premise edition built for your business

A fully customizable AI infrastructure, deployed on cloud or your servers with everything you love about Supervisely, plus advanced security, control, and support.

Start 30 days free trial➔
  • Maximum security: hosted behind firewall on your servers with advanced governance and privacy settings
  • Effortless integrations: single sign-on with LDAP or OpenID, cloud storage in AWS or Azure and powerful API & SDK
  • Priority support: dedicated slack chat, guided onboarding and personalized training sessions with experts
supervisely install
> downloading pre-requirements...
> pulling docker images...
> installing software...
> Done! Supervisely is running on port :80

supervisely update
> checking for updates...
> Your version is up to date!
Supervisely slack logo
#support
Team Manager 17:33
Hello @supervisely! Is there a way to create a project via API?
Supervisely 17:35
Sure thing, check out this docs!

Here’s why our customers trust us

BMW customer testimonial

BMW Group is using the Supervise.ly solution to create automated verifications for ensuring a very high product quality across the whole production chain in vehicle and vehicle component manufacturing.

BMW Group uses Supervise.ly to annotate manufacturing images from production lines in their world-wide plants for enhancing quality inspections using deep learning. The Supervise.ly tooling also supports the process for continuously updating AI models using semi-automated labeling. Supervise.ly is integrated into the BMW Group AI Platform in order to empower computer vision based AI use cases.

Engie customer testimonial
Quote mark

We use Supervisely since 2019. The key advantage of this tool is that Supervisely provides a complete data treatment pipeline. An important advantage is that a Supervisely instance can be deployed autonomously on a Client infrastructure, and distributed on different servers.

It helps to treat enterprise’s internal and often confidential data in a secured way. Together with a user-friendly interface, a clear documentation and a friendly and reactive support team it helps us to do Data Scientist work better and faster.”

engie-logo
Dmitriy Slutskiy
PhD Research Engineer
Resson customer testimonial
Quote mark

We originally set out to look for tools that could help us with data annotation, and we discovered that Supervisely excels at that and much more. It has become an integral part of our workflow in annotation, model training, and evaluation.

We've been exceedingly impressed with the customer support, addition of new features, and the flexibility of the publicly available SDK/API. The Supervisely team has also been fast to respond to support questions, and has shown a lot of openness when given feedback on potential improvements.

resson-logo
Travis Prosser
Engineering Specialist

Trusted by Fortune 500. Used by 65,000+ companies and researchers worldwide

Contact Us

Ready to get started?

Speak with people who are on the same page with you. An actual data scientist will:

  • Show live demo
  • Go through the concepts
  • Learn your case
  • Offer a tailored solution

Get you data labeled

Get accurate training data on scale with expert annotators, ML-assisted tools, dedicated project manager and the leading labeling platform.

Order workforce