Digitalize offline space with AI

Powerful image processingAI technology

mAy-I's image processing AI optimized for
security cam environments and visitor analytics

daram

Length of security cam footage analyzed so far

0hours

Approximately 410 years 11 months

Position and posture estimation

The key joint points of the visitors allow youto make specific estimates of their position and posture.

detail.

Accurate recognition and measurement

By tracking only each visitor, you canaccurately recognize and measure them as a visitor.

exact.

Reliable visitor tracking

Re-identification allows you to reliably continuetracking visitors, even in situations where tracking mightotherwise be interrupted due to overlap.

stable.

Accurate gender/age prediction even when wearing a mask

Because it detects visitors based on full-body images,it can accurately predict their gender/age range,even if part of their face is covered by a mask.

strong.

Digitalize offline space with AI. Digitalize offline space with AI. Digitalize offline space with AI. Digitalize offline space with AI. Digitalize offline space with AI. Digitalize offline space with AI. Digitalize offline space with AI. Digitalize offline space with AI. Digitalize offline space with AI.

Technology makes
offline space smart.

Feature

Sophisticated and precise image processing AI technology

Detection & Tracking

Finds people in each video frame and tracks them.

Re-Identification

Re-identifies whether a person appearing in different cameras is the same person.

Gender & Age Estimation

G&A Estimation

Extracts key features such as gender and age from faces and bodies.

Differences

An AI engine that analyzes in any environment
The high accuracy of AI 'daram'

Unlike Amazon Rekognition, which treats a person as a single object, 'daram' also detects key joint points of a person.

This enables additional information collection and more accurate detection of people who are partially overlapping or occluded.

This is because mAy-I's image processing AI 'daram' is an AI optimized for 'CCTV environments' and 'visitor analytics.'

Amazon Rekognition

‘daram’ in mAy-I

[Treat people as one and the same object]

[Detect major human joint points]

Recognize the same person once tracked, even if they overlap or are obscured

Recognize people as different if they are overlapped or obscured during tracking

Process

Building AI optimized for spaces.

mAy-I strives to maintain CGA accuracy of 95% across all spaces, and in practice shows approximately 90-98% accuracy depending on the environment.

* CGA stands for Counting (foot traffic and zone visitor counts), Gender (visitor gender), and Age (visitor age group).

Pre-compartmentalize areas in store and detect if a visitor crosses a zone or line.

Dwell time 10 minutes

Entrance

Help Desk

Exit

Doorways

Create a de-identified object for the visitor, then extract the gender and age data.

Dwell time 2 minutes

Tables

Process

Even at this very moment, our AI is getting stronger

mAy-I's image processing AI 'daram' continues to learn from collected CCTV footage data, and has analyzed over 3,600,000 hours of footage to date, becoming ever more powerful.

3,600,000

News

Already recognized mAy-I's technology

Top 5% paper selected at a world-class AI conference

K-startup innovation selected as 'Top 5% paper' at a world-class AI conference

mAy-I, which operates the image processing AI solution 'mAsh', announced that its paper on 'Research on Camera Bias in Person Re-identification AI Models' was published at ICLR 2025, one of the most prestigious AI academic conferences. ICLR is a leading academic conference in global AI research, ranked alongside NeurIPS and ICML as one of the top 3 AI conferences worldwide. mAy-I's paper was selected as a 'Spotlight' presentation, given to the top 5% of submissions.

March 6, 2025

Read article

Contact

'CCTV-based visitor analytics' is
what mAy-I does best.

Get the service brief

Digitalize offline space with AI

Powerful image

processing

AI technology

mAy-I's image processing AI optimized for security cam environments and visitor analytics

Length of security cam footage analyzed so far

hours

Approximately 410 years and 11 months

daram

Position and posture estimation

The key joint points of the visitors allow you to make specific estimates of their position and posture.

Accurate recognition and measurement

By tracking only each visitor, you can accurately recognize and measure them as a visitor.

Reliable visitor tracking

Re-identification allows you to reliably continue tracking visitors, even in situations where tracking might otherwise be interrupted due to overlap.

Accurate gender/age prediction even when wearing a mask

Because it detects visitors based on full-body images,it can accurately predict their gender/age range,even if part of their face is covered by a mask.

detail.

exact.

stable.

strong.

Technology makes offline space smart.

Feature

Advanced and sophisticated image processing AI technology

Detect and track objects

Detection & Tracking

Track people in each frame of security cam footage

Re-identifying objects

Re-Identification

Re-identify if the person appearing in the different cameras is the same person.

Gender & Age Estimation

G&A Estimation

Extract key features (gender, age, etc.) from a person's face and body.

Differences

High accuracy of AI engine ‘daram’ for analytics in any environment

Unlike Amazon Rekognition, which treats a person as a single object, ‘daram’ also detects the major joint points of a person.

This allows you to gather additional information and more accurately detect people who are partially overlapped or occluded.

mAy-I 's image processing AI engine ‘daram’ is optimized for 'security cam environment' and 'visitor analysis'.

mAy-I's ‘daram’

[Detects key joint points of people]

Recognizes the same person even when overlapping or occluded

Amazon Rekognition

[Treats a person as a single object]

Recognizes as different person when overlapping or occluded

Process

Create spatially optimized AI.

mAy-I strives to maintain 95% CGA accuracy in all spaces, and in practice it's 90-98% accurate, depending on the space and environment.

* CGA stands for Counting (# of foot traffic and space/zone visitors), Gender, and Age(age range), respectively.

Pre-compartmentalize areas in store and detect if a visitor crosses a zone or line.

Entrance

Dwell time 10 minutes

Exit

Doorways

Dwell time 2 minutes

Tables

Create a de-identified object for the visitor, then extract the gender and age data.

Consulting

Process

AI is getting more powerful, even now

mAy-I ’s AI is becoming more powerful as it learns more security cam footage data as it is collected, with a total of over 3,600,000 hours of analysis to date.

3,600,000

Media

mAy-I’s Technology Recognized Widely

Paper on K-Startup’s Innovative Technology Selected as ‘Top 5%’ at World-Class AI Conference

[Money Today] mAy-I, a startup which operates mAsh, a video processing AI solution, announced that its paper ‘Exploring the Camera bias of Person Re-identification’ has been selected for ICLR 2025, a prestigious AI academic conference....

March 6th, 2025

Read article

Contact

Visitor analytics based on security cam footage is best done by mAy-I.

Get the service brief

Powerful image processingAI technology

daram

Position and posture estimation

Accurate recognition and measurement

Reliable visitor tracking

Accurate gender/age prediction even when wearing a mask

Sophisticated and precise image processing AI technology

Detection & Tracking

Re-Identification

Gender & Age Estimation

An AI engine that analyzes in any environmentThe high accuracy of AI 'daram'

Building AI optimized for spaces.

Even at this very moment, our AI is getting stronger

Already recognized mAy-I's technology

'CCTV-based visitor analytics' iswhat mAy-I does best.

daram

An AI engine that analyzes in any environment
The high accuracy of AI 'daram'

'CCTV-based visitor analytics' is
what mAy-I does best.