Published on February 1, 2023
–
Updated on February 8, 2023
Dates
on the February 10, 2023
10am
Location
WAICF - Palais des Festivals, Cannes - 3IA BOOTH A13
Discover the results of the AI & Companies Week during the World AI Cannes Festival! PhD students will present their solutions after 1 week of work on a topic provided by an industrial partner.
The topics of this edition are the following:
- Centre d’Expérimentations Pratiques et de réception de l’Aéronautique Navale – Marine Nationale (Ministry of Defense)
- Title: Automated analysis of physiological data from VR aerial simulation session
- Abstract: From oculometric (eye movements) and cardiac data from a virtual reality headset the goal is to interpret the behaviour (under cognitive load, in cognitive overload or in nominal cognitive load) of the pilot performing the flight simulation or/and determine in which phase of flight he/she is (landing, take-off, simple transit, aerobatics). For this purpose, several data sets will be provided. In addition to these anonymised data sets, the researchers will have access to the phases of flight corresponding to each data item. They will also know at each moment what state the pilot is in. The team will have access to the computer with its simulator and the VR headset (Reverb G2 Omnicept) during the Hackaton week. The team can be partially supported by CSIO flight person.
- Caranx Medical is about the convergence between: Robotics, Imaging and AI.
- Title: Processing and segmentation of femoral artery ultrasound images
- Abstract: Automatic segmentation of medical images is an important step for deployment of surgical robots. For instance automatic detection of the location of the femoral artery from ultrasound images will allow the surgical robot to adapt in real time the trajectory of the probe to the patient’s anatomy. The goal of the project is based on segmented training images to propose an automated segmentation approach for new images. In order to increase the size the data set the team will also dispose of data from anatomic synthetic model thus also asking the question of domain adaptation of the model from synthetic data to true patients data.
- inHEART is committed to delivering the world’s most sophisticated, AI-enabled, digital twin of the heart to advance the care of patients living with cardiac disease.
- Title: Machine learning-based segmentation of medical images with partially labelled training data
- Abstract: The success of machine learning (ML) approaches, especially artificial neural networks, relies heavily on the volume of available data and the quality of its labeling. Unfortunately, labeling data is a time and money-consuming task; in the medical field it often requires domain-specific knowledge for the annotators. Is it possible to include partially or non-annotated data in a training dataset.