Yumain designs and deploys AI solutions
Maintenance detection security
Signal processing solutions with customized AI!
Our know-how
NEURAL TECHNOLOGY AND SIGNAL RECOGNITION TASKS
The overall signal recognition concept contains 3 strategic elements:
A perception sensor
(Camera, microphone, and many other types of sensors adapted to application contexts: EEG, ECG, magnetic field, vibration, etc.)
Mixed-use architecture,
VonNeumann (or Hardvard)
for data acquisition, information sorting and processing, and decision making
Inputs / outputs,
to learn the system and enable easy interaction with external systems
Nutritional analysis
Yumain helps with nutritional analysis at Ehpad and CHU, which are looking for a real-time diet management solution, enabling nutritional diagnosis and rapid adaptation of patient menus by offering a Deep Learning-based food recognition solution that not only identifies foods but also estimates their ingested quantity.
Face blurring
Yumain is helping to carry out real-time, in-vehicle integration processing aimed at blurring the faces of individuals present in the images taken in order to comply with the CNIL's RGPD by developing an in-vehicle image processing solution offering automatic face blurring, thus guaranteeing privacy protection and compliance with standards.
Character reading
Yumain helps automatic sorting in the tire recycling sector by developing a solution based on image recognition algorithms via artificial intelligence, using images from a camera in front of which the tires pass.
Counting elements in public transport
Yumain helps to count passengers with strollers or wheelchairs on buses and streetcars by offering an automatic recognition solution, based on deep learning, which enables the use of existing cameras. What's more, the results are directly analyzed and the images used are destroyed after in-house analysis, thus avoiding any problems linked to data management and RGPD.
Gas sensor
Yumain is helping to find a cost-effective tool for measuring ammonia concentration in complex environments, such as Comté cheese maturing cellars. This has led to collaboration with the Carnot Interdisciplinary Laboratory in Burgundy, and the development of a Proof Of Concept (POC) on the creation of a matrix of gas sensors with AI.
Level crossing surveillance
Yumain helps rail operators to ensure the safety of users around their level crossings by directly detecting the slightest sign of danger around the rails, then alerting users. Yumain also helps to ensure the reliability of existing safety systems by detecting the slightest malfunction and alerting maintenance staff for proactive, predictive intervention. What's more, the solution is easy to install, and can be integrated into existing infrastructures without the need to cut off track access.
Autonomous driving
Yumain is helping to improve the perception of autonomous vehicles by exploiting the event-driven camera, capable of detecting changes in illumination, beneficial for the detection of fast-moving objects and in difficult lighting conditions by proposing a solution based on its expertise that enables the processing of still or moving images and aims to provide three-dimensional perception for the localization and assessment of potential hazards in the environment of autonomous vehicles. Translated with DeepL.com (free version)
Catenary lift measurement
Yumain helps rail operators and maintenance managers to effectively measure catenary uplift by offering a mobile solution that is easy to transport and install on site, enabling reliable measurements to be taken quickly and without having to block sections of track to carry out the measurements. What's more, the measurements taken can be centralized and easily consulted by maintenance managers.
Digitizing paper plans
Yumain has developed a character, synoptic and graphic recognition solution to simplify the digitization and digitization of plans by developing an expert system capable of autonomously recognizing railway objects on geographical maps, reading documentary references and equipment tables, and then rendering the whole in a digital format, such as Excel.
Glove wear analysis
Yumain is helping to find a solution for real-time analysis, on an industrial site, of the type and wear of gloves used, in order to alert users to the suitability of gloves for their workstations, and to excessive wear, by developing, as part of a Proof of Concept (POC) project, various neural networks to automatically read the glove reference, and to check the level of soiling, thus alerting users to unsuitable or excessively worn gloves.
Surface analysis
Yumain is helping to find a solution to guarantee continuous quality control of surface condition, a task usually carried out by operators, by developing an operational demonstrator that has been installed on site.
Vibration analysis
Yumain is helping to set up a solution for analyzing parasitic vibrations in industrial equipment with a view to implementing predictive maintenance, by proposing an approach that transforms the vibration signal into a spectrogram, enabling real-time processing via artificial intelligence without the need for traditional signal processing technologies.
Electrical signal analysis
Yumain is helping to modernize railroad line and infrastructure management practices by increasing the digitization of data, particularly for turnouts, with a solution that leverages its expertise in signal processing and neural network structures, such as wavelet transform and autoencoders.
TROP![]()
HÉE DE L'EMBARQUÉ
On January 19, 2021, the E.C.S. Product with its pantograph detection and analysis application won us the Embedded Trophy in the Industry & Services IoT category.
This award was presented to us by CEA-List, a partner who has been with us since the beginning and recognizes Yumain's skills in embedded AI.
