The first Lab Aqsone Keynote took place on July 1, in the presence of our customers. The opportunity for us to present the Lab's activities and its achievements over the last three months.
The Aqsone Lab is a dedicated time for our employees to explore, innovate and learn about issues that are important to us such as the environment, health or social issues but also to master the latest technologies that address them.
How does data science make it possible to model and simulate the effects of drugs?
The first subject discussed during this keynote is a health project. The objective is to predict the effects of a drug on pathologies based on its molecular compound. This would speed up the drug discovery process.
To do this, Nicolas Cheifetz shows you how to approach this subject using machine learning and detailed analysis of undocumented data:
Over a period of three months, employees worked in teams on different projects, the results of which are shown below.
Deep Learning to protect the Great Barrier Reef
The first subject discussed during the Keynote gives an example of the use of Data Science and Computer Vision techniques in the service of the environment and biodiversity. The goal here was to detect in images the presence of a species of proliferating starfish endangering the balance of the Great Barrier Reef.
In this video, Hugo Naya shows you how to boost object detection algorithms such as YOLO, using Image Augmentation methods.
Audio signal processing applied to the detection of bird species
The second subject traces the participation of our teams in the 2022 edition of the BirdClef Challenge.
A challenge under the theme of biodiversity, which seeks data science solutions to classify bird species using recordings of their song.
Léa Besnard and Nicolas CHEIFETZ explain to you the main stages of audio signal processing and the results obtained in the challenge using the convolutional neural network (CNN).
Combining deep learning techniques to automate the Bechdel test
Our teams studied the automation of the Bechdel test which measures female (under)representation in cinema and audiovisual.
To do this, they combined computer vision techniques for detecting the gender of characters on screen with speech-to-text techniques for scripting scene dialogues and finally NLP (Natural Language Processing) techniques to analyze the topics covered in the dialogues.
Matthieu Vinette and Vincent Gargasson present their results to you in this video:
Reinforcement learning to optimize the energy consumption of buildings
The exceptional temperatures of this summer of 2022 have brought to the forefront our need to optimize our energy consumption, both in terms of cooling (air conditioning) and heating. Our teams have developed a reinforcement learning model to optimize the operation of the heating and air conditioning system of an office building. The model seeks a balance between consumption and occupant comfort.
Evan Graine and Colombe Becquart present the results to you in this video:
Machine learning for oral health
Dental health is a problem that everyone faces during their life. In collaboration with the RESTORE laboratory, our teams have designed a risk prediction model of tooth loss using machine learning techniques. Clustering techniques have been used in particular to highlight the characteristics of groups of patients at risk.
Adrien Schaffner presents his work to you in the video below:
Conclusion
The Aqsone teams continue to strengthen their expertise by working on projects around the themes of the environment, biodiversity, health and social issues.
The next Keynote will take place at the end of 2022.
A must see
