How Indoor Robotics Improved Training Data To Enhance Model Performance

How Indoor Robotics Improved Training Data To Enhance Model PerformanceAbstract background shapes
Table of Contents
Talk to an Expert

The Overview

Launched in 2018, Indoor Robotics is an innovative security company that combines humans and technology to bring more efficiency to surveillance. The company provides security and monitoring for indoor environments, including offices and warehouses.

Indoor Robotics helps clients who want to augment their security teams’ capabilities and scope. It also aids them in improving the overall management of their properties, providing round-the-clock inspections and collecting data on the state of their facilities. Indoor Robotics’ system captures data on events like leaks, fires, power outages, and unusual temperatures.

{{testimonial-1}}

AI-based Security Efficiency

Indoor Robotics’ security solution uses a self-navigating security drone, Tando™, that leverages AI to detect anomalies, perform inspections, and alert users to the presence of intruders, security breaches, safety concerns, and potential hazards.

These drones have fisheye lenses that provide full 360° camera coverage and can be directed remotely by clicking on a point on a map, leaving no blind spots, said Naty Shemer, R&D Group Manager at Indoor Robotics.

Indoor Robotics is able to signal atypical situations because drone flights follow the same path over and over again, during which they gather detailed intelligence that informs detection, says Shemer. And, he points out, nobody wants to watch hours of security footage to find unusual events. The IR system pings only the important things, minimizing security labor.

When a Tando drone detects an open door that’s supposed to be closed, a warning will be triggered and send alerts via The Control Bridge, the company’s platform, to the security team and/or relevant parties.

The Need For High Quality Data Annotations

But for Indoor Robotics to do that well, and win the trust of their clients, it must have a very high level of model accuracy. To accomplish this goal, they recognized the necessity of obtaining high-quality annotations to ensure continuous and consistent detection accuracy of their machine learning models.

“A good model that is able to perform detections is critical,” said Shemer.

They turned to Sama’s scalable annotation solution which was backed by SamaAssure™ the industry’s highest quality guarantee that routinely delivers a 99% client acceptance rate.

Sama’s dedicated workforce with deep domain knowledge provided the insights and quality needed to improve Indoor Robotics’ machine learning models and hit their milestones. That meant they were able to realize a healthy time to value, minimizing loss of time and avoiding overuse of resources.

Increasing Model Accuracy With Better Training Data

Doors are simple but also complex. A door is usually a rectangular structure lacking any distinctive features. They can also be different sizes so a ML model can’t even rely on a set of dimensions. And that makes it difficult for models to predict doors correctly, a potential problem considering that Tando drones regularly fly past and through doors.

Sama was able to enhance the quality and accuracy of the training data, including detection metrics, minimizing false positives and negatives, and used their Human-in-the-Loop process to label complete elements. Indoor Robotics was especially impressed with Sama’s tagging quality, which underwent an intensive quality assurance (QA) and AI model monitoring process.

“We significantly improved our training data, enhancing our object detection algorithm to identify people or doors,” said Shemer.

Launching a new ML model is a complex undertaking, and establishing a trusted security business that melds the best of technology and humans is particularly so. With Sama’s scalable annotation solution, Indoor Robotics accumulated sufficient accurately annotated data to train their models effectively. Sama met all annotation deadlines, even across complex images, enabling Indoor Robotics to meet launch deadlines, fulfill customer expectations, and start generating revenue quickly.

Maintaining Data Security

Indoor Robotics takes data security extremely seriously and must be compliant with GDPR (and other) regulations. Customers may provide some data, but the model can’t be trained using customer data alone, Shemer points out, due to privacy issues. By using Sama, which has its own workforce and does not rely on outside firms or contractors, the company is confident that it is operating within GDPR data regulations.

The Impact

  1. Enhanced model accuracy with improved training data. Sama’s proprietary Human In the Loop (HITL) approach provided high-quality annotations and insights to improve training data and enhance model performance.
  2. Met tight annotation deadlines to launch product on-time. Annotating high amounts of security footage for model training enabled Indoor Robotics to meet launch deadlines, fulfill customer expectations, and generate revenue.
  3. Maintained compliance with GDPR and other data security regulations. Sama’s exclusively in-house annotation workforce follows stringent security protocols in order to keep data safe and maintain compliance.

{{testimonial-2}}

We significantly improved our training data, enhancing our object detection algorithm to identify people or doors.

Naty Shemer
Naty Shemer
R&D Group Manager
at
Indoor Robotics

Sama is interested in building collaborative relationships, fostering a sense of being an integral part of our team rather than just another transactional outsourcing company. Additionally, Sama's in-house workforce is well-trained in annotating visual data, so they came up to speed quickly, providing invaluable insights to improve our ML models.

Naty Shemer
Naty Shemer
R&D Group Manager
at
Indoor Robotics
RESOURCES

Related Case Studies

Orbisk is Using Accurate AI to Help Restaurants Reduce Food Waste Up to 70%
CASE STUDY
MIN READ

Orbisk is Using Accurate AI to Help Restaurants Reduce Food Waste Up to 70%

Learn More
Swift Cost-Effectively Scaled Annotations Without Compromising Accuracy
CASE STUDY
MIN READ

Swift Cost-Effectively Scaled Annotations Without Compromising Accuracy

Learn More
Accurate Data Labeling Powers the Volumental Shoe Sizing App
CASE STUDY
MIN READ

Accurate Data Labeling Powers the Volumental Shoe Sizing App

Learn More
High-Quality Labels Power Accurate Search for Walmart’s 385 Million Online Visitors
CASE STUDY
MIN READ

High-Quality Labels Power Accurate Search for Walmart’s 385 Million Online Visitors

Learn More