Data labeling isn’t exactly a walk in the park. It’s a laborious, highly specified process that determines the success of any AI, ML, or NLP model. To bring some harmony to this typically dynamic process, we work tirelessly to hire and train a diverse group of experts that can label data en masse while also optimizing and scaling our clients’ workflows.
This article will outline the tools, approaches, talent, pricing, and scaling options we offer our clients at iMerit.
Who is iMerit?
iMerit is a tech-propelled solution for data labeling and annotation projects. Our tool-agnostic approach allows us to adapt to the unique data & tooling environments of our clients. Suppose you have data that’s partially labeled and done with the specificity of a particular tool. In that case, we’ll plug right into it, compliment the client’s workflow, and accelerate it accordingly to finish the project in the same vein it began.
iMerit has in-house tools for annotation, workflow management, and quality control. We adopt a mix-and-match approach with our proprietary tools, partner tools, and customer tools. Our Solution Architects put together the optimal tooling workflow for each project in consultation with technology, delivery, and Learning & Development.
iMerit: The Tools, The Data, & The Talent
1. iMerit leverages a flexible combination of tools and platforms to meet the needs of our clients
iMerit’s proprietary annotation tools are designed based on research and feedback from our skilled annotator workforce.Our technology team actively identifies areas within the annotation workflow where intelligent automation and deep learning can be leveraged. Semi-supervised automation has been enabled for several Computer Vision tasks by integrating APIs with the annotation workflows.
2. Client and Partner Tools
Many of iMerit’s clients have their own tools. That’s why our teams are hand-picked based on their experience with the widest variety of tools possible. iMerit also works with a robust ecosystem of tooling partners, including AWS Sagemaker, Appen, Dataloop, MightyAI, UnderstandAI, Deepen, SuperAnnotate, and more, which allows us to cater to any project.
3. Workforce Management and Skilling Tools
iMerit’s technology hub spans across all our operations. Our workforce operations are actively managed across skilling, rostering, and delivery using our tasking platform – the iMerit People Platform (iMPP). It takes care of skill mapping, skill development, workforce allocation, tracking, and performance monitoring.
4. iMerit Automated Tools
iMerit has in-house automated tools that prepare us to build any customized tool should the need arise. Our technology team actively identifies areas within the annotation workflow where intelligent automation and deep learning can be leveraged.
Data & Annotation Types
We are both tool and data-inclusive. iMerit has expertise in the areas of landmark annotation, bounding boxes, polygon & polyline annotation, semantic segmentation, LiDAR, 3D cuboid annotation, keypoint/landmark annotation, panoptic segmentation, and so much more.
For more information on our expertise and what each field entails, please view the following table:
Solution | Description |
---|---|
Computer Vision | |
Bounding Box Annotation | Mark rectangular boxes to identify targeted objects |
Semantic Segmentation | Annotate Images after segmenting into component parts |
3D Cuboid Annotation | Use 3D cuboids to illustrate desired objects by judging camera perspective, to obtain spatial-visual models |
Keypoint/ Landmark Annotation | Determine the shape changes by multiple consecutive points |
Polygon Annotation | Annotate object’s exact edges, regardless of shape |
Polyline Annotation | Mark lines to define pedestrian crosswalks, single lanes, double lanes, etc. for road recognition |
3D Point Cloud Annotation | Annotate 3D point clouds to support LiDAR and Radar |
Object Tracking | Locate and track objects across a series of images or point clouds in sequential frames |
Instance Segmentation | Identify each object instance of each pixel for every known object within an image |
Panoptic Segmentation | Couple instance and semantic segmentation |
Multi-Sensor Fusion | Combine LiDAR, Infrared and images from multiple angles captured from different sensors |
Natural Language Processing | |
Entity Recognition and Entity Classification | Identify, classify, and link relevant text strings for in-cabin automotive speech systems |
Intent Analysis | Analyze and categorize text into intents to enable Natural Language Understanding for in-cabin automotive speech systems |
Case Study: Leading Autonomous Vehicle Manufacturer
Use case: Segment data into predetermined classes of objects at the pixel level to train the ML algorithm
Source data: Car Footage
Data labeling type: Semantic Segmentation
Volume: Segmented more than 500,000 images
Outcome: Our quality experts ensured over 95% accuracy and the company was able to optimize its machine learning algorithm by significant margins.
Talent & Personnel
As previously mentioned, iMerit has 4000+ highly skilled full-time employees across the US, India, and Bhutan.
Our teams work with autonomous vehicle data engagements and undergo extensive training based on the geography, roads, and laws of the United States. This means that our teams are explicitly familiar with US road rules and road signs to ensure all autonomous vehicle data labeling is conducted adequately.
At iMerit, members of the full-time data workforce are recruited and skilled to perform a range of data annotation tasks that require different degrees of specialization and subject matter expertise. Our robust skilling structure helps produce both labeling specialists and domain experts to the benefit of our clients.
Quality Control & Security
iMerit labelers are trained for different initiatives, including work from home training, domain-specific certification programs, tools and applications training, and soft skills training. A closed feedback cycle is built-in with real-time monitoring and service delivery insights. We also employ an evaluation model that continually assesses deliverables, key metrics, quality control processes, and business outcomes.
iMerit is SOC 2 compliant, ISO 27001:2013 certified, and has been audited based on AICPA guidelines. We manage over 4,000 full-time employees across the US, India, and Bhutan under strict NDA. For high-security work, we operate dedicated and monitored facilities with strict security protocols.
- Access control: Physical access restricted through biometric readers and access cards and virtual access through Active Directory Services (ADS), and limited project access.
- Network security: Next-generation firewalls, a demilitarized zone for servers, IPS for protection from attacks, MAC-based Wi-Fi.
- Periodic reviews: User rights, vulnerability assessment testing, penetration tests.
- Risk management: Central patch management system, encrypted data backup and restore checks, two-factor authentication, hardening of end-user nodes.
Pricing & Scaling
Our pricing is based on the needs of our clients. We offer a full-time equivalent rate where cost is based on hourly pay per technician and a transactional rate where the price is composed by dividing the hourly rate by the projected tasks per hour.
Naturally, our clients don’t want to depend on us for everything. We are ready to support scale, when, and how the client needs it. This includes adding in extra hours to working on weekends to spinning up quick teams for ramp-ups.
- Our Expertise – We have done so much work within autonomous technology and other domains that technicians have become experts, and hence new workflows take less time to ramp up.
- Involvement of Learning and Development – We have our in-house trainers who take the project training forward using the materials shared. In this way, training new people during ramp-ups is managed, with limited time required from the client.
- Buffer resources – We provide more than the resources asked by the client and keep them at the same level as the others by rotating within the team. That way, the client does not lose any hours despite absenteeism.
As the data training process is highly iterative and constantly evolving, we at iMerit are committed to remaining as agile, versatile, and flexible as our clients need us to be. Our dedicated full-time group of experts is prepared to tackle any fluctuations in data volume. We also employ an in-house skilling structure whereby our experts can train client teams to make them project-ready in record time.
For example, iMerit recently scaled up a client’s processes by training and adding over 400 Technicians to the team in two months, bringing the team to over 900 data annotators. The knowledge of various tools and edge cases that iMerit acquires is a massive contributor to successfully scaling projects without compromising data quality.
Conclusion
When it comes to data annotation, flexibility is vital. iMerit’s proprietary approach to data annotation is designed with flexibility in mind. Our pricing is meant to meet each client’s unique needs, and our versatile approach to data labeling is intended to assure our clients that no combination of data, no matter how mismatched it may be, is too great of a challenge for us to take on.