AI & Data Operations

The data comes from the facilities using our platform. As these are all based in companies in the EHS field, the equipments, movements, objects, machinery used are different versions of the same products. This enables our AI models to get more accurate faster.

  • World’s #1 AI-Powered EHS Platform
  • Alerts Inspected by EHS Data Operation Specialists
  • Off-the-Shelf EHS Specific AI Models
  • Labeling of Real-Time Collected Data
  • Transparent AI Accuracy Evaluation by EHS Specialists
  • World’s #1 AI-Powered EHS Platform
  • Alerts Inspected by EHS Data Operation Specialists
  • Off-the-Shelf EHS Specific AI Models
  • Labeling of Real-Time Collected Data
  • Transparent AI Accuracy Evaluation by EHS Specialists
  • World’s #1 AI-Powered EHS Platform
  • Alerts Inspected by EHS Data Operation Specialists
  • Off-the-Shelf EHS Specific AI Models
  • Labeling of Real-Time Collected Data
  • Transparent AI Accuracy Evaluation by EHS Specialists
  • World’s #1 AI-Powered EHS Platform
  • Alerts Inspected by EHS Data Operation Specialists
  • Off-the-Shelf EHS Specific AI Models
  • Labeling of Real-Time Collected Data
  • Transparent AI Accuracy Evaluation by EHS Specialists
  • World’s #1 AI-Powered EHS Platform
  • Alerts Inspected by EHS Data Operation Specialists
  • Off-the-Shelf EHS Specific AI Models
  • Labeling of Real-Time Collected Data
  • Transparent AI Accuracy Evaluation by EHS Specialists
  • World’s #1 AI-Powered EHS Platform
  • Alerts Inspected by EHS Data Operation Specialists
  • Off-the-Shelf EHS Specific AI Models
  • Labeling of Real-Time Collected Data
  • Transparent AI Accuracy Evaluation by EHS Specialists
Data Collection
01
325 Million images are processed every second by our AI algorithms.
02
The collected data comes from the real use-case scenarios of the facilities using our platform. Thus, the data pool includes the best examples for the related use-case development.
03
Different versions of the same equipment type are used in various facilities (such as helmets, reflective vests, and forklifts in different colors and designs). In this way, a more enlarged data pool is obtained compared to any other platform, and training the best AI models in the EHS field becomes possible.
EHS-related high performing AI models
70+ object categories, including personal protective equipment, indoor/outdoor vehicles, crane heads, etc.
In-house labeling structure by our Data Operations Specialists via a tailor-made labeling tool MAIA
Homogeneous data distribution with regular checks for better and more generalized models
The labeling process organized according to the customers’ needs
AI Models and Algorithms
AI Models
  • 10x Object Detection
  • 2D & 3D Body Pose Estimation
  • Proprietary Person Reidentification AI
  • Face Anonymization
  • Fire & Smoke Detection
  • Housekeeping Models (Leakage and Spill, Unattended Object Detection, Clean Vehicle Roads, Clean Pedestrian Ways)
Tracker

Receive camera stream and analyze each use-case with the prediction from AI models and trigger alert if a violation occurs.

Ergonomy Solutions

The posture of the human body is detected with a body pose estimation model, which in turn is used to decide whether an action is performed in an ergonomically correct way.

Object Detection

The object detection model is responsible for detecting objects in the frames. It receives a frame as input and outputs bounding boxes and names for any object of interest on that frame. Afterward, these predictions are used to analyze each use case.

Accuracy Check

After your custom use cases are set up on the Intenseye platform, if non-compliance is detected, you will receive an alert.

To ensure the accuracy of alerts, we take samples and evaluate them as true alerts or false alerts. This process increases the accuracy performance of our models.

92%+ average accuracy of the alerts for the facilities using our platform. (in the last quarter of 2021)

Periodical accuracy checks of models inspected by Intenseye EHS Data Operations Specialists.

Sharing the insight offered by the manual inspections and the results of checks with deep learning engineers to increase accuracy rates.

Rearranged labeling roadmap according to model performances.

Completed reports shared by customers transparently.

Tailor-Made Tools

The success of our platform is supported by our custom-developed Maia and Maat tools. The collected and labeled images from live cameras are integrated into our system and the accuracy rates of the alerts are calculated.

Maia is our AI-powered image annotation tool that we built internally. Maia receives video streams from cameras and automatically labels the video feed. Our labeling experts add missing annotations afterward.

Get demo

Maat is our accuracy check platform. Our team of EHS experts measures our accuracy and transparently provides the results on a continuous basis.

Get a demo

Success Stories

Increasing the accuracy rates of Housekeeping Models at one of our customer’s facilities’

01

EHS Data Operation Specialists inspected the accuracy of the housekeeping models in the first two weeks

02

The labeling process has been completed in a week by the Data Operation Specialists on the images collected from the cameras with the Housekeeping rules.

03

The research on a new model is completed by the engineers

04

The labeling process has been completed in a week by the Data Operation Specialist on the images collected from the cameras with the housekeeping rules

05

With the deployment of the new model, the average accuracy rate of the use case increased from 66% to 98%, an increase of 32%

footer-1@2x
footer-2@2x
Let’s grow together
Looking for great people who share our passion! Help us create a meaningful impact on workplace health and safety globally.
footer-3@2x
footer-4@2x
Schedule a Demo