How to control growth phases intelligently and efficiently

Just a little while ago, it was hardly possible to gather such information based on camera technology. Now, all these new vision applications are made possible by artificial intelligence. AI itself is no easy technology – there are usually lots of steps required from gathering training images to a fully and independently working application. Users have to manage the data collection, the data set labelling, the AI knowledge, the programming of the neural network, the training, and validation and finally the hardware platform for the AI processing to run on.

_{Text continues underneath image}

Easy technology for all

The all-in-one AI solution IDS NXT ocean takes care of everything between the user knowledge and physical deployment – this makes the technology available to anybody. With the help of the IDS NXT lighthouse cloud software, even users without prior knowledge of artificial intelligence or camera programming can train a neural network. They do not have to set up their own development environment but can start training their own neural network right away. This involves 3 basic steps: To upload sample images, to label the images (e.g. “good” / “bad”) and then to start the fully automatic training. The trained network can then be executed directly on the IDS NXT industrial cameras, turning them into powerful inference cameras ready to solve the required task.

_{Text continues underneath image}

AI-based image processing

Another advantage of IDS NXT is the ability to perform AI-based image processing in an edge device. AI processing can be pretty intensive since neural networks can sometimes have millions of parameters, which means that processing and memory resources must be carefully managed. IDS has developed an FPGA AI accelerator core and worked hard to optimise the neural networks created with IDS NXT lighthouse to ensure that AI processing runs efficiently on IDS NXT cameras. There is no need to execute the AI on an industrial PC, since it is already integrated in the cameras. This paves the way for image processing tasks such as the classification of fruit varieties or the detection of unusual characteristics, shapes or colours, which can be directly solved by the camera.