AI Superior GmbH

Traditional deal-sourcing methods are time-consuming and limited by industry-standard classification systems like NACE (used in the European Union). These systems lack codes for emerging market fields such as machine learning and artificial intelligence, making it difficult to identify relevant companies within these market fields.

To solve this, we developed a solution that addresses the challenges faced by investors in finding niche markets and assessing their financial states efficiently. Powered by Natural Language Processing techniques and Deep Learning models, our solution collects, processes, analyzes, and displays data from various sources, enabling users to perform semantic and syntactic searches, explore company clusters, review similar companies, and more. The solution also incorporates financial data extraction and aggregation, providing insights into the development of relevant companies over time

Our client, an ophthalmology center, required an automated solution that could assess eyelid positioning and identify proptosis-related abnormalities.

To overcome this challenge, we used deep learning algorithms to create Eye Metrics, a cross-platform web application, accessible on mobile (iOS and Android) and desktop devices, that empowers users to capture facial images and calculates key metrics for ophthalmological analysis. 

Eye Metrics revolutionizes eye health analysis by accurately assessing eyelid positioning, detecting abnormalities related to proptosis, and providing a seamless experience for healthcare professionals.

Our customer was looking for a way to reward good driving skills by offering cheaper insurance rates. However, accurately assessing driving behaviours and providing fair discounts based on individual performance has been a challenge. Traditional methods rely on limited data and subjective assessments, leading to inconsistent results. To overcome these challenges, AI Superior set out to develop an advanced system that could process and analyze telematic data to accurately evaluate driving behaviours and calculate personalized discounts.

We leveraged deep learning algorithms to develop a model capable of analyzing various telematic data collected from the driver’s phone. The system could analyze data from accelerometers, gyroscopes, GPS, and other sensors to detect driving modes, such as whether the user was driving or a passenger, and track driving behaviours like exceeding the speed limit, unnecessary acceleration, abrupt hard braking, and more. A scoring algorithm was then applied to each trip to provide an overall score. Based on this score, a personal discount was calculated for each driver. Additionally, the system provided recommendations to promote safe driving practices, further incentivizing drivers to improve their performance. 

Our customer was faced with the issue of graffiti that appears overnight on their city walls. Locating these graffiti spots and organising their removal can be quite challenging and time- consuming. On top of that, manual methods of detection are often inefficient and can lead to delays in the graffiti removal process.

To tackle this, we developed an advanced model that enables rapid, real-time detection and localization of graffiti. Using a vehicle equipped with a 360-degree camera to patrol the city streets, our solution applies deep learning algorithms to detect graffiti with high accuracy and capture photographs of it in real time. Each detected graffiti spot is accompanied by the GPS location of the vehicle at the time of detection. This information is then transferred to an interactive map with markers. Each marker identifies a graffiti location on the map, enabling users to see the exact location and 360-degree image of the graffiti. Relevant metrics, such as the total number of graffiti spots, graffiti density map, and average size of graffiti, can also be extracted from this platform.

Our customer, an Ophthalmology Centre, faced the challenge of accurately estimating the volume of fat and muscle in the human eye. This information was crucial for monitoring eye health, evaluating the effectiveness of interventions, and conducting comparative analyses. 

We employed deep learning techniques to develop an advanced model capable of accurately segmenting fat and muscle tissue in each slice of an MRI orbit scan. By leveraging the model’s capabilities, we achieved highly precise volume estimation for the eye’s anatomical structures. Our deep learning model was designed to handle different views of the eye, including Coronal, Sagittal, and Axial, ensuring comprehensive analysis of the orbital region. 

Cities face numerous challenges in maintaining and inspecting their road infrastructure, particularly when it comes to efficiently detecting and evaluating road damage, such as potholes. Manual inspections are time-consuming, subjective, and often result in delays in identifying and repairing road defects. 

To address this challenge, we developed a platform that utilizes deep learning segmentation models to accurately detect and evaluate potholes and road damage. The platform accepts video footage or individual frames as input and applies a deep learning model to accurately segment potholes. Key features such as size and area are extracted for each segmented pothole, providing essential information for estimating severity levels. The platform also includes a scalable GIS application for visualizing road damage, customizable notifications for critical potholes, and filtering capabilities to prioritize repairs. Plus, it can be integrated into a real-time video processing pipeline for continuous monitoring and instant detection.

Our client, a solar energy company, faced the challenge of time-consuming manual methods for analyzing residential roofs for solar panel installations. To improve overall efficiency and accuracy, we developed an automated solution that enables efficient planning for solar panel implementation. The solution utilises deep learning models to accurately detect and segment roofs into distinct areas, and estimates the area and dimensions of each segment, enabling optimal solar panel placement on residential roofs

A leading real estate online platform required a solution that would facilitate the value assessment of urban zones within a city. To provide industry professionals with more accurate insights, we developed a segmentation model to assist in this process. 

We develop a solution that facilitates the assessment of prices for different areas within a city. The project involved analyzing the city map using a semantic segmentation approach to generate a detailed segmentation map. This map classified pixels into predefined classes such as roads, residential houses, infrastructure, green areas, and open land areas. Historical maps from different years were also incorporated to gain insights into the city's evolution over time, particularly focusing on the availability of open areas for real estate development. These insights allowed the real estate company to assess the value and desirability of specific regions within the city and make informed price evaluations based on their expertise and market knowledge.

Summary: For a large semi-government organization, our team developed a system that detected litter objects from images captured by drones and helped to manage trash collection activities. We designed and developed a GIS-based application that allowed convenient interaction with detected litter objects and facilitated trash collection activities via optimal route planning and tracking of the collection progress.

Challenge: Tons of litter objects were dragged by waves to a coastal area, scattering them across a territory of around one thousand square km. Being dispersed across many single islands these objects were quite difficult and costly to identify by a human. Hence, it was required to design and implement an efficient strategy that allowed the detection and collection of litter.

Solution by AI Superior: We applied our proprietary computer vision technology for object detection, classification, and segmentation and designed and developed an interactive GIS-based application to display results and operate collection activities. The computer vision detection technology was applied to images captured by a drone. The drone flights were operated by a partnering team that provided RGB images covering the whole territory (the flight altitude was around 50 metres high above the land)

Outcome and Implications: the developed solution allowed to reduce the time required to perform trash detection activities by the factor of 25. This  resulted in significant cost savings, halving the overall detection and collection costs. Furthermore, the collection time decreased by a factor of 4 while the automated computer vision solution allowed a 7% higher detection accuracy compared to a human expert. Additionally, this system decreased carbon footprint by a factor of 19.

The picture shows the dashboard of the application.

Summary: Our team developed a drone-based application that was able to detect and report 25 different types of construction debris. The solution allowed to automatize the construction site inspection process while reducing human involvement and average inspection costs considerably.

Challenge: A city municipality requested a solution that allowed them to fully automate construction site compliance monitoring and detect abandoned construction debris such as bricks, cement blocks, sand heaps, metal and wooden sticks, etc. It was crucial to have an automated solution that would minimize human involvement in the inspection process thus reducing labor costs as well as the time required for an inspection.

Solution by AI Superior: We applied our proprietary computer vision technology for object detection, classification, and segmentation to detect 25 different classes of construction debris. We built a GIS dashboard to allow selection of a construction site and the visualization of all the debris detected within it. Additionally, for every detected object the system provided an estimated size (area) of construction debris (for a single object and clusters of objects of the same type) as well as the amount of detected objects. The application provided insights, while employing a GIS dashboard and exposed APIs to query detection results – this allowed the solution to be integrated to practically any other system.

Outcome and Implications: The solution was adopted by multiple city municipalities demonstrating its operational and economical effectiveness. Furthermore, according to our customer estimates, the system saved 320 man-hours per month and reduced average inspection costs by 40%.

The picture shows the result of the Construction debris detection from a drone. The different objects are dyed in different colors.

For a niche insurance company operating in a medical/health domain we developed a prediction machine learning-based model to estimate the risk of an economical loss. The machine learning model is based on neural networks and built by consuming historical medical data over five consec-utive years. The developed model significantly outperformed statistical approaches. With this mod-el, the customer was able to optimize its pricing policies which resulted in significant savings.