The world of artificial intelligence (AI) can sometimes seem complex and inaccessible, but the Tygron Platform proves that this does not have to be the case. With a focus on open source and accessibility, Tygron enables users to develop, train and apply AI models themselves. In this blog, we show how easy it is to get started with AI and how Tygron supports this.
Why open source is important
Tygron strongly believes in sharing knowledge and technology. That is why the neural networks and training data they develop are published open source. This means that everyone has access to the tools and data to create their own AI models. By using open standards such as ONNX (Open Neural Network Exchange), users can easily integrate their own trained models into the Tygron Platform. This promotes innovation and collaboration, both within and outside the AI community.
How accessible is AI on the Tygron Platform?
Building and applying your own AI model consists of 3 steps:
- Creating your training data;
- Training your model;
- Applying your model;
1. To train your model, you need a GIS program, such as QGIS. If your goal is to recognize objects on satellite images, you can export a high-quality satellite map via the Tygron Platform. You can then trace the objects on the satellite map in your GIS program. When you have created sufficient training data, you load it into Tygron. You can then export this training data from the platform in a way that it can be directly loaded as training data.
2. In order to train your model, you will need software in addition to your training data. The easiest way to do this is to use the Anaconda software environment. In Anaconda, you can work with Python-sctips and Jypyter-lab.
You can download our step-by-step plan via GitHub. In fact, this is a repository with scripts and datasets that you need to train a neural network (GitHub – Tygron/tygron-ai-suite). Based on the step-by-step plan (How to train your own AI model for an Inference Overlay – Tygron Preview Support Wiki), you can train your own AI model. The type of AI model that you train is an RCNN model.
You can then export your model and save it as an ONNX file.
3. The final step is to run your AI model. To do this, you need to upload the saved ONNX file to the Tygron Platform. With the new AI Inference Overlay, you can connect your AI model to maps that you want to apply your model to. For example, if that is the satellite map, connect this map to the AI Inference Overlay. (Inference Overlay – Tygron Preview Support Wiki). The Inference Overlay will now run your AI model.
This step-by-step plan is suitable for both beginners and experts. In the process of training the model and running the model, beginners can use the default settings and the expert in turn has enough parameters to play with to make the trained model better.
Practical applications of AI
The possibilities of AI on the Tygron Platform are broad. A practical example is the use of an AI model for the Vallei en Veluwe water board to improve water management on the Veluwe. We generated an AI model to generate detailed maps of vegetation and soil structures based on satellite images. This not only showed that traditional maps are not accurate, but this method also provides much more accurate data that provides better insight into infiltration capacity and microrelief. This helps in creating simulations and ultimately in taking the most targeted measures (see also Hoe gebruiken we AI om neerslag beter vast te houden waar het valt?).
Some other examples are:
- Improving tree data, for example by supplementing tree data from public sources with data on private land.
- Improving geographic information on waterways and structures in the registers of water boards.
- Identifying zebra crossings and walking routes to better map mobility.
These applications show that AI is not only theoretical, but also directly contributes to practical solutions.
Collaboration with educational institutions
Tygron works closely with colleges and universities to spread knowledge about AI. Students are given free licenses to experiment with the platform, which teaches them how to create, train and apply neural networks. This contributes to the development of a new generation of experts who are familiar with AI technology.
Conclusion
AI doesn’t have to be magical or complicated. The Tygron Platform offers an accessible way to get started with this technology yourself. Thanks to the open source approach and practical tools, everyone can experiment, learn and contribute to innovative solutions. So what are you waiting for? Dive into the world of AI and discover how you can improve your data and strengthen your analyses!
The functionality described here is available on our Preview Server.
- If you are interested in getting started on the Preview Server yourself, request access by sending a message via the contact page on this site or by sending an email to support@tygron.com
- Also check out the available documentation such as AI Suite, Demo Iteration Project and our training offer Tygron Academy
- Do you have any questions? We are happy to help you via support@tygron.com.
