obss business landing


With our technical and academic knowledge in the fields of artificial intelligence (AI) and machine learning (ML), which direct the technologies of the future, we develop innovative solutions that provide the most accurate answers to the needs of the corporations. With the comprehensive work we carry out based on the available data, we automate the business processes of the corporations we work with, thus contributing to increase their efficiency even more! We customize our partners’ technological transformation processes to the future by providing end-to-end services in all of the following areas:

• Computer Vision
• Data Science
• Natural Language Processing (NLP)
• Simulation and Synthetic Data Generation

We make success a part of our business processes!
As the Machine Learning & Artificial Intelligence team, we prioritize vertical specialization with both our academic and engineering perspectives. With the help of our wide range of software and architectural team, which consists of expert engineers who have received doctoral education at reputable universities in Turkey, we are able to implement, smoothly and efficiently, the ML and AI products which have higher risks compared to standard software products. In this way, we make it possible to gain the necessary know-how and experience in business processes long before the operation phase. 
By providing service to our customers with the most advanced technologies, we make it possible for them to gain speed on the way to the future!
While we enable operations such as object detection, image segmentation and exposure estimation to be performed on the screens with computer vision technology, we are also able to create realistic and interactive models or environments through simulation and synthetic data generation. In addition, by means of data science, we enable the implementation of works such as forecasting, modeling and statistical analysis by creating systems that can be manually adapted according to work situations. With our competency in natural language processing (NLP), we lead the way in improvements of services such as voice assistant, text analysis and chatbot in various languages, especially in Turkish and English.

Areas of Expertise

With the solutions we provide in the field of computer vision, we have experience in all areas such as object detection, object localization, image segmentation, classification, exposure estimation and tracking on real images. We develop models that can detect small objects within the large-size images over the GPU memory, allowing them to be manually adapted according to the needs in the field. Among our prominent projects to date and offered as an open source, we also have an artificial intelligence video analysis service named SIVA which was developed for the computer vision library SAHI (Slicing Aided Hyper Interface) and the Sprongo ski/sports platform. With the autonomous sea vehicle Piribot, which we have developed as a part of our R & D studies, a more effective coastal detection can be achieved with the image segmentation models we use.

For the areas where it is not entirely possible to obtain sufficient and high-quality data, we provide the opportunity to gain realistic experiences by using synthetic data generation and simulation technologies. In this way, we create a qualified training and preliminary preparation opportunities for building systems and operation field studies, by supporting the increase of field-dominance. In addition to being able to carry out object recognition and sample segmentation studies in widely used data formats such as COCO, we can also simulate outputs such as point cloud data or radar tracks with the additional modules we use. We can create interactive and kinematic models on objects by setting up different types of sensors and cameras for this purpose.

We offer state-of-the-art solutions in the field of natural language processing (NPL) in order to develop services such as voice assistant and chatbots in various languages, mainly Turkish and English. We carry out data collection and cleaning processes by conducting our studies on neural network-based models or classic models. In addition, we are also able to obtain more advance outputs by using the web browsing technology. In this direction, we ensure that the needs are met in many areas, including text classification, sensitivity analysis, intent analysis, question detection, automatic question creation and the speaking feature. With our competency in this field, we can also obtain effective results in Turkish which is a rich language, in terms of morphologically and linguistically, due to its agglutinative nature.

In addition to the advanced machine learning techniques developed in order to extract value from existing data, we carry out field-specific estimation, detection, optimization, analysis and modeling studies by using field information and statistical methods. We provide the opportunity to obtain the best results on the relevant work areas by developing models that can easily be manually adapted for potential problems, intended use and targets. Due to the emphasis we place on the analytical skills and learning abilities of our colleagues while establishing our team, we are able to develop effective solutions so that the various needs in different sectors and business lines can be fulfilled. Along with obtaining realistic and consistent outputs on issues such as sales, credit and supply-demand forecasting, we are also diversifying our services according to the specific areas such as the troubleshooting, malfunctions and malfunction locations of the radar systems.

Sample Projects

As the Machine Learning and Artificial Intelligence team, we carry out, in-depth R&D and prototype studies, especially on autonomous vehicles. We aim to develop innovative solutions for military and civilian naval applications with the Piribot autonomous sea vehicle which is one of our most prominent projects in this field.
Providing its own energy through solar panels, Piribot is able to perform physical and chemical analyzes for current conditions with its sensor package that can be configured according to specific purposes. In addition, it offers the opportunity to be able to make detailed explorations on the sea and the sea coast through computer vision. With its machine learning and internet technology of the objects, Piribot also has the ability to act as a fleet or a swarm, independent of human factor.

We are among the stakeholders of the MARTERA (Maritime and Marine Technologies for a New Era) project carried out by the European Union within the scope of the Horizon 2020 Program. As one of the four companies participating in the ADRIATIC program, which includes Norway and Romania, we undertake the mission planning and simulation works of the autonomous marine vehicles.

Carried out under the umbrella of Era-Net Cofund, MarTERA EU is a technology-oriented project for the protection of the seas and for the development of environmentalist maritime activities. The project also focuses on critical international issues such as maritime transportation, oil and gas activities, and security. The OBSS Machine Learning and AI team is in the ADRIATIC branch of the project and is actively carrying out works in the development of sensor, automation and tracking technologies in the maritime field.

As OBSS, we have the status of being the technology partner of the Sprongo platform which provides IT-based services for sports such as skiing, swimming, golf, rowing and gymnastics. Provided by the platform, we are developing the artificial intelligence video analysis services called SIVA (Sprongo Intelligence Video Analytics) though computer vision technology, based on exposure estimation. In this way, we provide support for the athletes and related organizations to be able to carry out advanced studies on posture and technique and to improve their performance.

The SIVA platform anatomically analyzes the moves of the athletes via the videos obtained, allowing them to be examined in detailed graphics. The platform, which provides numerical data on posture position by detecting the joint points, enables the determination of moves and postures that need to be improved by making comparisons with the athletes that have a perfect technique. It also allows adding technical evaluations to videos with the opportunity to draw and record sound on visuals.

We offer the SAHI Computer Vision Library, which we have developed, as an open source service. Compared to similar applications, we enable more detailed analyzes and extraction to be carried out in object detection, object localization and segmentation with the advanced technologies we use in this field.

Object detection and classification, which are one of the most important branches of computer vision technologies, can still be performed in a limited way in today’s conditions. On the other hand, SAHI enables the end-to-end identification of even the smallest objects on the large-scaled images. For this, the application can label all the objects on the image by using pixels and colors. Furthermore, the last version of the application can easily be installed via the “pip” software which is the package management system of Python.

For more details…
SAHI Computer Vision Library
SAHİ: Görsel Kütüphane

With our dominance in data science, we carry out studies on malfunction detection and localization of radars which are important in the fields of military and civil security. We perform the necessary analyzes and create models through the obtained structured, unstructured and semi-structured data. Thus, we ensure that the problem is detected without any loss of extra time and effort.

Although radars, which are important for strategic areas, have durable equipment, they still may experience problems in terms of functioning under certain conditions. The conditions that cause the malfunctioning of the radars can be detected and analyzed in detail with the help of the available data. In this direction, virtual tests are carried out on the radar according to the increase and decrease of factors such as heat, humidity, wind, etc. As a result, the outputs needed to identify the problem and to strengthen the necessary equipment can be obtained.

Creating meaningful texts in different languages is one of the most challenging topics in the area of natural language processing. As the OBSS Machine Learning and Artificial Intelligence team, we conduct innovative studies on the subject of creating a question which is one of the less studied sub-branches in this field! By identifying the expressions that may be meaningful to ask in Turkish and English texts, we carry out modeling and software development studies in order to turn them into appropriate question sentences.

With the help of the system we have developed for natural language processing, we also ensure that the questions can be created flexibly when the desired topics to be asked or the clues that form the questions are given by the user. With the automation we provide in the question preparation processes, we offer advanced opportunities to the education world in the following areas:

•Reducing the question preparation costs,
•Gathering attention with the quizzes that can be prepared instantly,
•Increasing the general comprehension level of the students, etc.

As for the Jury library, which allows easy calculation, at one time, of many metrics that we have created, we are sharing it with the natural language processing community as an open source in order to evaluate the quality of the texts produced:
Jury Open Source Library
Jury Açık Kaynak Kütüphanesi

The use of cameras, instead of specialized sensors for drone detection, promises a serious potential in the field of image processing due to the cost advantage and the flexibility it provides. However, the difficulty of detecting the drones through the cameras is one of the most important problems that still need to be overcome in this field. In drone use, the diversity of the environment, the size of the drones in images (especially in remote detection scenarios), the difficulty of distinguishing between birds and drones and the scarcity of the significant data on this issue are the main reasons why this problem in question cannot be overcome.

Regardless of the fact that the drone recognition technology with cameras is complex, its high potential and the opportunities it offers, make this issue a noteworthy business area in the software field. Based on this, as the OBSS Machine Learning and Artificial Intelligence team, we are also working on the development of this technology in the field of image processing. And we are able to offer our partners comprehensive solutions in the following areas:
Training of the best models for detection,
Synthetic data generation under different conditions in order to produce high-performance and durable systems,
Enrichment of datasets
Making classifications with significant filters through the determinations, etc.