Idiap Research Institute

Idiap Research Institute
Name	Idiap Research Institute
Established	1991
Location	Martigny, Valais, Switzerland
Director	John-Marc Odobez
Focus	Artificial intelligence, speech processing, computer vision, machine learning
Affiliations	École Polytechnique Fédérale de Lausanne, University of Geneva

Idiap Research Institute is an independent research centre founded in 1991 in Martigny, Valais, Switzerland concentrating on artificial intelligence, speech processing, computer vision, and machine learning. The institute operates within a Swiss and international network of academic, industrial, and governmental partners and has contributed to projects that intersect with initiatives at institutions such as École Polytechnique Fédérale de Lausanne, University of Geneva, Swiss National Science Foundation, European Research Council, and CERN. Researchers at the institute publish with venues like NeurIPS, ICML, CVPR, ACL, and INTERSPEECH and collaborate with companies including Google, Microsoft, IBM, Amazon (company), and Facebook.

History

Idiap was founded in 1991 in Martigny following Swiss cantonal support and early connections to École Polytechnique Fédérale de Lausanne and the University of Geneva. Throughout the 1990s it developed links with projects supported by the Swiss National Science Foundation and participated in European Framework Programmes associated with Horizon 2020 and the Framework Programme 6. In the 2000s the institute expanded research groups drawn from alumni of Stanford University, Massachusetts Institute of Technology, Carnegie Mellon University, University of Cambridge, and University of Oxford, and engaged with technology transfer partners such as IBM Research and Siemens. In subsequent decades Idiap researchers led or contributed to collaborative grants from the European Research Council and consortiums involving ETH Zurich, Imperial College London, Technical University of Munich, and Max Planck Society.

Research Areas

The institute focuses on multidisciplinary topics including speech and audio processing, natural language technologies, computer vision, multimodal interaction, and machine learning methods. In speech and audio labs, work relates to automatic speech recognition, speaker diarization, and speech synthesis, connecting to communities around INTERSPEECH, ICASSP, HTK, and Kaldi. Natural language and dialogue teams publish for ACL, EMNLP, NAACL, and tackle tasks related to machine translation with links to Europarl, WMT, and Moses (software). Computer vision efforts address object recognition, human pose estimation, and scene understanding with outputs presented at CVPR, ICCV, and ECCV, and tools such as OpenCV and TensorFlow are commonly used. Research into multimodal interaction and social signal processing interfaces with work on Affective computing, Human–computer interaction, CHI Conference on Human Factors in Computing Systems, and projects involving Microsoft Research and Google DeepMind.

Facilities and Resources

The institute maintains laboratories equipped for audio capture, motion capture, and high-performance computing, including GPU clusters supporting deep learning frameworks such as PyTorch and TensorFlow. It operates dedicated data annotation environments and corpora management influenced by standards from LDC (Linguistic Data Consortium), ELRA (European Language Resources Association), and participates in shared tasks like CHiME Speech Separation Challenge and AMIGOS (dataset). For vision and multimodal experiments, the institute employs motion-capture stages, depth sensors associated with Kinect, and eye-tracking systems referenced in studies at MIT Media Lab and Max Planck Institute for Informatics. Its compute infrastructure interconnects to Swiss national resources like Swiss National Supercomputing Centre and cloud services from Amazon Web Services and Google Cloud Platform.

Collaborations and Partnerships

Idiap maintains collaborations with academic partners including École Polytechnique Fédérale de Lausanne, University of Geneva, University of Lausanne, ETH Zurich, University of Cambridge, and Imperial College London. Industry partnerships have included firms such as Google, Microsoft, IBM, Amazon (company), Facebook, NVIDIA, and regional technology companies in the Swiss innovation ecosystem. The institute also participates in European consortia involving Horizon 2020, FP7, European Research Council projects, and transnational collaborations with Max Planck Society, Inria, CERN, and research groups at Carnegie Mellon University and Stanford University. Engagements extend to public-sector stakeholders in Valais alongside cultural institutions and health organisations that mirror collaborations seen with World Health Organization initiatives and clinical partners in academic hospitals.

Education and Training

Researchers at the institute supervise doctoral students registered with universities such as University of Geneva, École Polytechnique Fédérale de Lausanne, and University of Lausanne, contributing to doctoral programmes and postdoctoral training. Idiap hosts internships, summer schools, and workshops aligned with conferences like NeurIPS, ICML, CVPR, ACL, and supports teaching through guest lectures and joint courses with EPFL and regional universities. Training activities include hands-on sessions with frameworks referenced by PyTorch, TensorFlow, and software engineering practices promoted by GitHub and continuous integration approaches used at Travis CI and Jenkins.

Awards and Impact

Researchers associated with the institute have received recognitions from bodies such as the Swiss National Science Foundation, European Research Council, and prizes announced at conferences like NeurIPS and ICML. Scientific contributions have influenced standards and toolchains adopted by industrial labs at Google Research, Microsoft Research, and IBM Research, and have been cited in projects related to autonomous vehicles at companies like Tesla, Inc. and academic programs at CARLA (simulator). The institute’s datasets and methods have been reused in challenges such as CHiME and WMT, and staff have served on program committees for NeurIPS, ICML, and CVPR, shaping research agendas across speech, language, and vision communities.

Category:Research institutes in Switzerland