Summary
- Profile Type
- Technology offer
- POD Reference
- TOIT20240611016
- Term of Validity
- 11 June 2024 - 11 June 2025
- Company's Country
- Italy
- Type of partnership
- Investment agreement
- Research and development cooperation agreement
- Targeted Countries
- All countries
Contact the EEN partner nearest to you for more information.
Find my local partner
General information
- Short Summary
- An Italian ICT company has led the exploitation of lightweight AI algorithms for real-time denoising and image restoration, countering adverse weather and environmental conditions. This technology enhances object detection for first responders during rescue missions. The company seeks investors to support further development. This technology stems from an EU project on DRS-02-Technologies for first responders.
- Full Description
-
Severe weather and environmental conditions, such as rain, smoke, snow, haze, or low light can affect the visual perception of a human. This can be a serious obstacle for many different scenarios like search-and-rescue operations or autonomous driving since the detection of objects of interest and the navigation to possible destinations may become extremely tough and sometimes even dangerous. Adverse weather and environmental conditions degrade the visual quality of the images and consequently both decrease the situational awareness of the human or the robot and the performance of the AI-based object detection algorithms. Moreover, the adverse conditions can degrade the visual quality of the images and can reduce the performance of the algorithms, resulting in missed detections, misdetections, or noisy output. In addition, computer vision AI algorithms are often complex and computationally demanding, requiring high-end computers to run in real-time. Their deployment on edge devices can result on very slow responses or be altogether prohibitive. In search-and-rescue operations, for example, FRs often come across many stressful situations, experience extreme safety risks and having the additional obstacle of the weather may reduce their efficiency during a rescue mission and simultaneously increase the risk of injuring themselves. Lack of visibility is a problem for many people and also for machines, not only in work related environments but also on a day-to-day basis. Whether the lack of visibility is due to rain, haze, snow or even the lack of light, the fact is that it prevents the final user from fulfilling their objective efficiently. On top of this, this lack of visibility is not something you can always foresee since an unpredictable situation, so you won’t always have the appropriate tool in order to fix it. For example, a flashlight, an umbrella, fog light and so on. The effectiveness of the various types of sensors and cameras used in autonomous cars may also be influenced negatively by the heavy rain, snow, smoke, or low light, which may lead to an accident.
The proposed solution is a suite of lightweight AI denoising and image restoration/improvement algorithms able to counter different types of adverse weather and environmental conditions and perform in near-real-time, with the ultimate goal to aid a detection algorithm to show a better performance, in detecting objects that may be useful for the FRs during their rescue missions or enabling autonomous cars to drive in harsh weather conditions safer and more reliable.
The solution was tested and evaluated in pilots in a relevant environment and the involved universities and research centers are searching for investors to support the ongoing development and improvement of this technology. - Advantages and Innovations
-
The proposed solution is able to detect a wide variety of objects of interest even when the image contains noise from rain, snow, haze, smoke or darkness. Usually, the object detection algorithms are trained on images with clear weather conditions where the distribution of colors differs. Thus, the performance of the existing solutions proposed for this task is reduced when applied to cases where the weather is bad, or where there exists darkness or smoke.
Additionally, this solution has a lightweight architecture, which makes it easy to run in real-time on edge devices and avoid delays in response.
Finally, apart from some common objects that most of the existing solutions detect (e.g., person, car etc.), the proposed solution will have the ability to detect objects that best address FR’s needs (e.g., clothes, gloves or boots that may be dropped by victims).
More efficient computer vision AI, improved search and rescue, use-cases for visual AI expanded to imperfect conditions, improved performance for live image restoration on edge devices.
New and emerging technologies are relatively unknown outside of specialist or enthusiast circles. Bringing such technologies to use in the most important of tasks, i.e., saving lives, will contribute to broader societal impacts of improving the security perceptions of the wider public. Moreover, the resulting wider acceptance of new technologies will help pave the way for their application and use in diverse fields beyond crisis response, encourage European industry to invest in them, and ultimately align with the goals of the Digital Single Market. - Stage of Development
- Lab tested
- Sustainable Development Goals
- Goal 9: Industry, Innovation and Infrastructure
Partner Sought
- Expected Role of a Partner
- Looking for investors interested in supporting the future development of the technology.
- Type and Size of Partner
- R&D Institution
- University
- Type of partnership
- Investment agreement
- Research and development cooperation agreement
Dissemination
- Technology keywords
- 01003003 - Artificial Intelligence (AI)
- 01003012 - Imaging, Image Processing, Pattern Recognition
- Market keywords
- 05004002 - Rescue and emergency equipment
- 02007021 - Other Artificial intelligence related
- Targeted countries
- All countries