Proven R&D team continues system/software development activities in the following areas to be used in automation projects in the Proven’s business domain:
Data Augmentation and Synthetic Image Generation
Explainable AI
Federated Learning
Anomaly Detection Using AI
TestINT
TestINT offers a compact, specialized platform for image augmentation and AI testing, specifically designed for computer vision problems related to street-view and aerial images. Its image augmentation feature allows users to diversify test data sets within minutes, using a wide range of options. TESTINT can generate new data sets by:
- Altering weather conditions (adding rain, snow, fog),
- Converting day images to night,
- Modifying images based on the entered prompt,
- Placing realistic objects, and applying input corruptions and adversarial attacks.
These synthetic challenges help users save time, effort, and money, while make AI models safer. TESTINT’s testing functionality allows users to create test scenarios by choosing from a variety of model performance and test adequacy metrics. Comprehensive test reports are then prepared to help users identify potential risks in their AI models. The project has been supported by The Scientific and Technological Research Council of Turkey (TÜBİTAK), Industrial R&D Projects Grant Programme 1501.
XLARIFY
XLARIFY is an innovative, no-code Explainable Artificial Intelligence (XAI) platform designed to make AI model decisions transparent and understandable. XLARIFY offers a comprehensive suite of tools for analyzing image, text, and tabular data, helping users identify hidden patterns, biases, and the underlying logic of AI decisions.
We introduce Explainable AI Technology to various sectors for enhancing decision-making processes and improving model transparency.
- Analyze complex AI models across different data types.
- Identify and mitigate biases in data and models.
- Provide visual and textual explanations for AI decisions.
XLARIFY provides an opportunity for AI developers and users to understand and trust their AI models by offering detailed explanations and analyses. It supports various AI-driven industries such as healthcare, automotive, and finance by ensuring the reliability and transparency of AI decisions. The project has been supported by The Scientific and Technological Research Council of Turkey (TÜBİTAK), Industrial R&D Projects Grant Programme 1501.
FLOPS
In traditional machine learning methods, data is transferred to a central server for processing, whereas federated learning allows data to remain on distributed devices or resources. This approach ensures the secure processing of privacy-sensitive data. FLOPS is a no-code, web-based platform that enables various edge devices to collaborate on training machine learning models without sharing their local data. The platform features optimized algorithms to accelerate training and reduce communication costs through client selection, model compression, and federated aggregation. The project has been supported by The Scientific and Technological Research Council of Turkey (TÜBİTAK), Industrial R&D Projects Grant Programme 1501.
ProVER
In modern vehicles, various functions are managed by Electronic Control Units (ECUs), and communication between these units occurs over the in-vehicle network. As the number and complexity of vehicle functions increase, so does the number of ECUs, leading to more complex in-vehicle network message traffic. The increasing volume of data makes it challenging for test engineers to manually detect and diagnose anomalies by examining log files. Rule-based software programs also struggle to identify anomalies caused by complex root causes. To address these challenges, our team has been developing AI-based solutions for anomaly detection in in-vehicle networks. ProVER significantly speeds up the analysis process and eases diagnosing underlying causes of anomalies. The project has been supported by The Scientific and Technological Research Council of Turkey (TÜBİTAK), Industrial R&D Projects Grant Programme 1501.