6 minutes to read Computer vision Advances in deep learning have revolutionised the field of computer vision and unlocked applications that were unthinkable just one decade ago. From quality assurance in industry to medical imaging in hospitals, the automatic detection, understanding, and measurement of image and video content has broad applications. And it's positively impacting many industries already. Today's main computer vision challenges lie in developing fair and robust models that work reliably on diverse forms of data. That's why at Zühlke, we emphasise custom labelling and build interpretable and transparent models that are properly validated using high-quality data. Here are some core computer vision applications: 1. Computer vision in medical imaging The application of this machine learning technology in medical imaging is a highly promising development in global healthcare. For a Swiss MedTech company, which recently was awarded the CE certification for medical devices, we set up a medical machine learning process and also designed, implemented, and validated a regulatory compliant data platform and computer vision use cases. 2. Computer vision in animal conservation Computer vision can greatly increase the efficiency and scale of animal observation. Automatic species identification, and even individual animal tracking, allow us to improve the study of behavioural dynamics of entire populations. In turn, this enhances our understanding of ecosystems and how to improve biodiversity. 3. Computer vision in predictive maintenance Being able to detect and even pre-empt the wear of mechanical parts is an essential part of equipment maintenance. Together with a Swiss transportation company, we built a machine learning solution to automatically measure the wear of pantographs on passenger trains. Using computer vision segmentation of photographs of the trains, our algorithm can measure deterioration of materials. This allows early scheduling of maintenance and therefore minimises the downtime of rolling stock. Natural language processing Rapidly find the most relevant documents for a task, automatically summarise a long document, and automatically address your customers’ most pressing questions. It's all possible with a machine learning technology called natural language processing. NLP allows us to derive structured insights from unstructured written and spoken language. Deep learning based NLP can automatically classify documents, extract entities and their relationships from documents, as well as summarise and even generate text and images. Here's a closer look at some of the key NLP use cases: 1. NLP for electronic medical records Electronic medical records (EMR) contain vast amounts of information about a patient’s medical journey. Machine learning methods and EMR can be used to prevent disease and improve treatment decisions, which improves patient outcomes. For a Swiss healthcare provider, we developed cutting-edge recurrent (LSTM) deep learning models to classify EMR to support healthcare professionals and streamline hospital processes. 2. NLP for text classification Natural Language Processing can be used to automatically classify documents and take appropriate action. For a Swiss transportation company, we developed a transformer-based deep learning system for high-volume email processing. Our solution understands and classifies each email's topic, automatically assigns emails to appropriate agents, and analyses end client problems and trends over time. 3. NLP for question answering Question Answering (QA) systems allow for information retrieval from large bodies of knowledge, automatically answering questions posed in natural language. For a Swiss authority we developed a QA system based on state-of-the art, transformer-based, deep learning methods. With our solution, the public body can provide citizens with rapid and automated answers via a natural dialogue based interface. Time series Time series methods allow you to derive insights from data that unfolds over time. Think climate and weather data, medical data, financial data, and industrial data. This enables us to understand and forecast trends in complex areas like supply chains and urban planning. It allows us to detect and predict anomalies and generally classify time series – for instance, to predict circulatory failure in ICU patients. Here are some key use cases for time series methods: 1. Time series for climate change Climate change is a key challenge of our time, and addressing it with a broad spectrum of solutions will ensure the best possible outcome. Machine learning will be a key part of this solution, for example in improving energy production and use, optimising transportation and routing to reduce emissions, enhancing production to reduce waste, and monitoring the environment and the climate as a whole. 2. Time series for financial forecasting In the financial sector, quantitative methods play a key role. For a telecommunications provider we developed predictive time series models to perform financial performance forecasting on multiple time series, which supports the provider’s financial experts in their work. 3. Time series for renewable energy The transition to renewable energy will help mitigate climate change. For an energy provider we developed time series models to predict the risk of failure in wind turbines based on a large quantity of near real-time sensor signals, enabling their service partners to proactively perform maintenance, ensuring a high degree of uptime and optimal energy production levels. Regulated AI On average, a medical doctor in the US has only seven minutes’ consultation time per patient. By applying medical AI solutions, we can free up doctors' time from repetitive tasks, prevent misdiagnoses, and improve access to better treatment around the world. Thanks to our regulated AI process and our experts on quality assurance, we can build compliant machine learning models that are safe and validated for clinical use. 1. Regulated AI for compliance Helping general practitioners diagnose complex diseases can lead to faster and better treatment that would otherwise be missed due to underdiagnosis. Based on a machine learning model from the research department of a large pharmaceutical company, we developed a system to support physicians in their differential diagnosis for pulmonary diseases. To do this, we brought the model under design control to meet all regulatory requirements for a validation study, and to create the basis for FDA approval. 2. Intensive care with regulated AI Continuously monitoring a patient’s vital signs is an important step in preventing morbidity and mortality. Machine learning can help interpret these signals and give early warnings for high-risk events like sepsis or circulatory failure. 3. Regulated AI in FemTech FemTech aims to positively impact women’s health and focuses on digital health solutions and products in the area of fertility and period-tracking, pregnancy, sexual wellness, and menopause. We helped AVA Women to extend the functionality of their fertility tracking bracelet to further indications. This put the medical device wearable in a higher risk class. We reviewed and improved the processes used for developing the machine learning models. Thanks to this, AVA now fulfils the regulatory needs and successfully obtained FDA clearance of its 510(k) application. Explore machine learning technologies and use cases with Zühlke Data exchange and application is essential for realising your greatest opportunities and solving the biggest issues of our time. But complex data silos, poor data quality, and regulatory red tape can make this an impossible task. Our teams empower you to co-innovate better solutions together with diverse partners – based on a foundation of openness, transparency, accessibility, and effective governance. We develop and scale systems in a principledand sustainable way, and we're leaders in human-centred, interpretable, and responsible AI. Interested in what AI solutions could do for your organisation? Talk to us today about how to ideate, validate, and operationalise machine learning use cases in a principled and responsible way. You might also like... Data Strategy – Responsible AI: how to develop ethical AI applications Learn more Data Strategy – One step at a time: how to build the future of AI regulation Learn more Life Science and Pharmaceutical Industry, Medical Device and Healthcare AO Foundation: AI enables patient privacy breakthrough in clinical applications Learn more Contact person for other locations United Kingdom Switzerland Germany Singapore Contact person for United Kingdom Dan Klein Chief of AI & Data Dan is Chief of AI & Data and has extensive experience working across a diverse range of sectors, including government, transport, telecoms, and manufacturing. As a skilled engineer and strategic advisor, Dan effectively connects the needs of leadership with the technical expertise of teams to successfully drive data transformation initiatives for organisations. He brings a unique combination of strategic thinking and deep knowledge of data and engineering to his consulting work. Contact Daniel.Klein@zuhlke.com +44 207 113 5306 Your message to us You must have JavaScript enabled to use this form. First Name Surname Email Phone Message Send message Leave this field blank Your message to us Thank you for your message. Contact person for Switzerland Philipp Morf Head AI & Data Practice Dr. Philipp Morf holds a doctorate in engineering from the Swiss Federal Institute of Technology (ETH) and holds the position head of the Artificial Intelligence (AI) and Machine Learning (ML) Solutions division at Zühlke since 2015. As Director of the AI Solutions Centre, he designs effective AI/ML applications and is a sought-after speaker on AI topics in the area of applications and application trends. With his many years of experience as a consultant in innovation management, he bridges the gap between business, technology and the people who use AI. Contact philipp.morf@zuehlke.com +41 43 216 6588 Your message to us You must have JavaScript enabled to use this form. First Name Surname Email Phone Message Send message Leave this field blank Your message to us Thank you for your message. Contact person for Germany Tobias Joppe Director Customers Solutions Tobias Joppe studied automation and control engineering at the TU Braunschweig and was most recently head of a innovation team at Siemens AG. He has been with Zühlke since 2008, is a partner and, as Director Customers Solutions, is responsible for the Trend Lead Data Science in Germany. In his role, he builds the bridge between cutting-edge technology and current customer needs. Together with customers, he translates visions and goals into a strategic roadmap and concrete project procedures. As Director Customers Solutions, many completed interdisciplinary projects form the basis of his experience. Contact tobias.joppe@zuehlke.com +49 511 220 021 43 Your message to us You must have JavaScript enabled to use this form. First Name Surname Email Phone Message Send message Leave this field blank Your message to us Thank you for your message. Contact person for Singapore Nicolas Lai Business Development Manager APAC Nicolas supports the Healthcare market unit at Zühlke Asia, focusing on innovative digital strategy and product development initiatives with global and local customers. Nicolas is passionate about helping clients connect the dots, bridging the gap from conceptualisation to implementation. Contact nicolas.lai@zuhlke.com +65 8780 2998 Your message to us You must have JavaScript enabled to use this form. First Name Surname Email Phone Message Send message Leave this field blank Your message to us Thank you for your message.
Data Strategy – Hybrid companies – a step towards a new era of cooperation between humans and machines Learn more