NIPG / ELTE — the AI and informatics hub
The lab that builds and connects the AI. Every other technical capability in this ecosystem either originates at NIPG or passes through it.
Institutional context
The Neural Information Processing Group (NIPG) sits within the Department of Artificial Intelligence (DoAI) at the Faculty of Informatics of Eötvös Loránd University (ELTE), Budapest.
- DoAI focuses on human-centred, ethical and trustworthy AI. Founded in partnership with Robert Bosch Ltd., it merges academic research with practical industrial application, including deep learning and distributed intelligence, and aims to lead Hungary in computational intelligence and composite AI.
- The Faculty of Informatics, established 2003, is a leading Hungarian centre for computer science education and research, hosting nearly 3,000 students in Budapest and Szombathely, with extensive partnerships across multinational and domestic technology companies.
- ELTE is the oldest continuously operating and largest public research university in Hungary. Founded 1635, nine faculties, roughly 30,000 students. Closely linked to several Nobel laureates and world-renowned scientists.
Why NIPG is the hub
Every named technical thread in this ecosystem touches NIPG:
- Speech Gap’s inventor pool overlaps with NIPG-adjacent researchers at Szeged / HUN-REN.
- The ADOS-2 multimodal behaviour analysis work involves ELTE researchers together with Rush and Sorbonne.
- The avatar and rehabilitation work (AIRS) is an NIPG line.
- The DFKI collaboration — annotation, assistive interaction, human–robot perception — runs through NIPG. → Supporting collaborations
- The Richter / Colleyeder project originated as an ELTE-adjacent initiative.
- Argus Cognitive’s researchers largely came out of ELTE.
NIPG supplies the method layer every clinical use case needs: multimodal feature extraction, speech/face/gaze/blink processing, efficient multimodal transformers, label-efficient video segmentation, multi-view 3D pose, semantic 3D reconstruction, edge pose estimation, and composite-AI reasoning over these features.
What the group actually does
NIPG develops multimodal perception and composite-AI systems that convert speech, video, human movement and environmental observation into temporally structured and interpretable information for healthcare, rehabilitation, social interaction and robotics.
Capabilities with identifiable public implementations include multimodal feature extraction; speech, face, gaze and blink processing; multimodal transformers; label-efficient video segmentation; multi-view 3D pose; and research-oriented semantic robotic mapping. Others have been demonstrated through research prototypes or publications only.
The full 17-row capability table, with per-row evidence pointers and repository links, is at Capability portfolio.
Caution. Capability maturity varies widely, from public repository to “research prototype, code not verified.” This portfolio is not a single validated product suite; the per-row evidence pointer is what establishes readiness. See Capability readiness and safeguards.
Leadership
András Lőrincz — NIPG founder; Fellow of the European Association for Artificial Intelligence (EurAI); member of the European Laboratory for Learning and Intelligent Systems (ELLIS); convenor of the INNOFAIR 2026 workshop.
Role in the ecosystem
| Relationship | What NIPG contributes |
|---|---|
| Semmelweis / VIKOTE | The measurement and analysis method layer for a funded clinical trial |
| Speech Gap | The respiratory-signal extension and the software rebuild, layered on Szeged-origin IP |
| Argus Cognitive | Shared research lineage; the translation case study that shows where NIPG-style research can end up |
| Michael E. Benros | Methods that could be validated against large-scale Danish clinical and register environments |
| Erik Szabó | Research output that his five-test screen can evaluate for venture viability |
Related pages
András Lőrincz · Capability portfolio · Speech Gap · Semmelweis / VIKOTE · Ecosystem map · Editorial safeguards