INSIGHTS INTO THE NEW FEATURES OF HALCON 21.11.
17 Nov 2021 | Reading time: 2 min
Interview with Stephan Hahn
Portfolio Manager at STEMMER IMAGING
Our portfolio manager Stephan Hahn gives you some interesting insights into the new features of HALCON 21.11. and how these are enabling new applications in new markets.
The new MVTec release HALCON 21.11 Progress includes many new and improved features. Which markets do you think will benefit most from these features?
Stephan Hahn: The wide variety of features that HALCON offers open up an extensive range of possible uses in countless applications across all industries. However, one specific market that comes to my mind is in agriculture, which is seeing a rapid adoption of machine vision. In this area, I see great potential for HALCON and its new features.
In this market in particular, there are all kinds of possible applications. Can you give us a practical example that involves a specific feature?
Stephan Hahn: In this context, I’d like to highlight a new technology called instance segmentation which extends the functional scope of deep learning features in HALCON 21.11. This new feature combines semantic segmentation and object recognition and offers great advantages in supporting farming and harvesting. It’s especially beneficial when identifying and measuring naturally grown structures such as in agricultural produce, e.g. directly on the field, where organic objects are very close to each other, touch and even partly overlap, as objects can be assigned to different classes with pixel accuracy.
In your opinion, is it a big effort to make an existing machine fit for deep learning in a short time?
Stephan Hahn: No, no great effort at all. This is precisely what makes it so unique: Existing systems can be retrofitted with simple hardware such as a USB stick in combination with the OpenVINO toolkit. In this way, you can realise significantly higher speeds for deep learning inferences on Intel processors including CPUs, GPUs and VPUs. It couldn't be easier!