European Funds for Małopolska, Republic of Poland, Co-funded by the European Union, Małopolska – go to the About the project section
Wellms Rekomender 2.0

A module of the Wellms platform

See when your group’s attention drops

Rekomender 2.0 shows online instructors the group attention indicator at any given moment, suggests when to take a break, and after class points out the slides with the highest and lowest scores.

Book a free demo See the versions
Group attention over time Class report
Chart of the group attention indicator during a class, divided into topics and breaks: the indicator drops in the middle and recovers after a break
The attention indicator is dropping. Consider taking a break in the next few minutes.

Results are always aggregated for the whole group, with no view of individual participants.

How it works

No extra tools for participants. All you need is a class in Wellms
and cameras switched on by those who choose to take part in the analysis.

92.2% attention classifier accuracy
+1.09 pts higher class rating from participants
+1.66 pts higher group focus rating in the instructor survey
60 lessons in a controlled real-world trial

Participants join

Participants connect to the class in Wellms. Turning on the camera is voluntary.

The system analyses the video

Every few seconds the model detects participants’ attention. Analysing a single frame takes 0.23 s on average.

You get guidance

Results for the whole group appear on a live dashboard and in a post-class report,
down to individual slides.

Two versions, each suited to its use

Each version is configured to comply with the EU Artificial Intelligence Act (AI Act) for its intended users.

For education

Rekomender EDU

For universities, language schools, training companies and L&D teams that run live online classes.

  • Real-time group attention indicator
  • Break suggestions for the instructor
  • Post-class report, slide by slide
  • Comparisons across classes and courses
  • No emotion recognition and no assessment of instructors
Sign up for a pilot
For research and sales

Rekomender INSIGHT

For market research and UX agencies, and B2B marketing and sales teams that test content with audience groups.

  • Test group reactions, second by second
  • Report slide by slide or scene by scene
  • Forecast of overall reception
  • Comparison of two versions of the material (A/B)
  • Analysis only with the consent of the participants
Timeline of test group reactions across successive parts of the material Order your first test

Proven in real classes

In a controlled trial, 6 instructors each taught 5 online lessons without the system and 5 with it, with 50 participants in total. After each lesson, participants rated the class and instructors rated the group's focus.

We set the 3-minute threshold for break suggestions through interviews with educators who teach online.

Research and development results
MeasureResult
Class rating by participants
(scale 1–10)
+1.09 pts vs. classes without the system
Group focus in the instructor survey (scale 1–10)+1.66 pts vs. classes without the system
Attention classifier accuracy
(3-fold cross-validation)
92.2%
Reaction model accuracy
(INSIGHT, 7 categories)
76.9%
Break suggestion accuracyaverage error under 3 minutes
Training data1,216 surveys and over 15,000 labelled images

Privacy and AI Act compliance

Since 2 February 2025, the AI Act has prohibited emotion recognition in education and the workplace. We designed Rekomender 2.0 so that each version complies with the rules that apply to its users.

  • The EDU version uses a binary classifier that recognises only attention or its absence, with no emotion recognition.
  • Results are always aggregated, and instructors are never analysed.
  • The system does not run with fewer than 2 cameras or when fewer than 60% of cameras are on.
  • We process data on servers in the European Union under a GDPR-compliant data processing agreement.
  • In the INSIGHT version, every participant gives consent; we provide templates for information notices and consent forms.

Built on Wellms

Wellms is an e-learning platform released under the open MIT licence. Its core is free, and more than 20 training platforms have already been built on it. You pay only for Rekomender and the services you need.

Already using another platform? Rekomender can connect to other e-learning systems through an API.

Pricing

EDU pilot, 30 daysPLN 990
Rekomender EDUfrom PLN 390 / month
INSIGHT presentation testfrom PLN 1,490

Prices exclude VAT. We will send the full price list once you get in touch.

Ask for a quote

Who is behind it

ESCOLA website

ESCOLA S.A.

A software house building digital solutions for education, business and public administration, and the creator of the Wellms platform. Listed in the Deloitte Technology Fast 50 Central Europe in 2023.
www.escolasoft.com

AISEEMO website

The AISEEMO research team

We developed the image analysis algorithms with the research team of AISEEMO, a spin-off of the AGH University of Krakow specialising in image processing and emotion recognition.
www.aiseemo.com

See Rekomender
in action

Sign up for a free, 90-minute online demo session.

biuro@escola.pl · +48 798 896 837

European Funds for Małopolska, Republic of Poland, Co-funded by the European Union, Małopolska

About the project

Development of the “Rekomender 2.0” system using artificial intelligence to improve the quality and effectiveness of online teaching

Project objective

To improve the quality and effectiveness of online teaching by developing and deploying a system that uses artificial intelligence to give instructors real-time information about their group's attention and reactions.

Target groups

Universities, language schools, training companies and L&D teams that run online classes, the instructors who teach them, and their participants.

Results

A new product, the Rekomender 2.0 system, developed to technology readiness level TRL 8 and brought to market as a module of the Wellms platform.

Beneficiary
ESCOLA S.A.
Programme
European Funds for Małopolska 2021–2027, Measure 1.1 “Research and development projects of enterprises”
Project number
FEMP.01.01-IP.01-0298/23
Project period
11.2024–01.2027
Total project value
PLN 5,430,710.55
European Funds contribution
PLN 3,180,825.80

Project tasks

  1. Task 1 · R&D Developing the training data format and a data collection environment, and labelling the data
  2. Task 2 · R&D Training an attention recognition model
  3. Task 3 · R&D Training an emotion recognition model
  4. Task 4 · R&D Studying attention and emotion patterns during classes
  5. Task 5 · R&D Implementing the models in a SaaS service
  6. Task 6 · R&D Developing a system prototype
  7. Task 7 · R&D Further developing the system prototype
  8. Task 8 · deployment Bringing the product to market: the
    Rekomender 2.0 system