What is the project
Sustainable processing for computer vision ensuring compliance with privacy guidelines is a research and development initiative led by IST, IST-ID and INESC-ID. It focuses on creating optimized computer vision and artificial intelligence (AI) algorithms that reduce computational requirements, execution costs and energy consumption, while ensuring privacy and regulatory compliance. The project supports the integration of optimized AI into autonomous retail solutions under the PT-Smart Retail Agenda.
Project at a glance
Focus: Sustainable and privacy-compliant computer vision and AI for autonomous retail
Key demonstrators: VICAN large-scale camera calibration; human and head pose estimation under occlusion; AI-on-demand platform; GDPR (General Data Protection Regulation) pre-validation software; cloud execution of algorithms on sensitive data using trusted hardware
Operation: Edge computing environments with optimized algorithms and secure execution mechanisms
Objectives
- Optimize computer vision algorithms to reduce computing time, cost and energy consumption.
- Migrate and optimize algorithms for edge computing environments.
- Develop secure and anonymous computing mechanisms with low processing overhead.
- Test, adapt and fine-tune edge platforms for AI, computer vision and sensor data processing.
- Ensure compliance with the General Data Protection Regulation (GDPR).
What we delivered
Development of optimized computer vision algorithms, including VICAN (Very Efficient Calibration Algorithm for Large Camera Networks) for large-scale camera calibration and human/head pose estimation methods robust to occlusions; creation of virtual supermarket environments to evaluate algorithms while addressing GDPR constraints; deployment of an AI-on-demand platform with a redesigned interface, integrated object detection, tracking and 3D reconstruction modules, and automated pipeline orchestration for edge computing; implementation of software for automatic GDPR pre-validation with mechanisms for verifying data collection, storage and processing; research outputs including scientific publications, reports, conference presentations, seminars and supervised MSc and PhD theses.
-
Metrics / KPIs
MSc thesIs (IST / IST-ID)
Value: 10
Notes: Listed individually (2025)
MSc thesIs (IST / INESC-ID)
Value: 22
Notes: 2023-2025
PhD thesIs (IST / INESC-ID)
Value: 2
Notes: 2024
Journal articles (IST / IST-ID)
Value: 6
Notes: 2023-2025
Conference & workshop papers (IST / IST-ID)
Value: 8
Notes: 2023-2025
Reports (IST / IST-ID)
Value: 4
Notes: 2026
Research scholarships (INESC-ID)
Value: 14
Publications (INESC-ID)
Value: 52
Training activities & trainees (INESC-ID)
Value: 24
Notes: 2 training activities + 22 graduated MSc
Events for dissemination (INESC-ID)
Value: 46
Notes: Fairs, exhibitions, conferences
Jobs created (INESC-ID, PRR-related)
Value: 4
Summary
This project advances sustainable and privacy-aware computer vision for autonomous retail. It combines algorithm optimization, edge deployment and secure execution with regulatory compliance mechanisms aligned with GDPR. Key outcomes include large-scale camera calibration (VICAN), occlusion-robust pose estimation, an AI-on-demand platform with automated pipeline orchestration, and GDPR pre-validation software. The initiative is supported by extensive scientific output, supervised research, and broad dissemination activities.
.webp)
