Menu

Innovations of Deep Learning in Agriculture Sector

While AI is still early in its development, it is already being recognized to be an important and rapidly growing application with expected substantial impacts on our societies. These pioneering AI algorithms are enabled through several high performance computer systems that are challenging designers on many fronts. The traditional data center designs are rapidly migrating from general-purpose CPU-only solutions towards combinations of CPUs and GPUs or TPUs, bringing new and more stringent demands on design of server power solutions.

 

Infineon offers industry’s highest efficiency power stages that utilize Infineon’s market leading OptiMOS™ MOSFET technology. Through continued Infineon advancements in its power semiconductor technology, our devices are becoming increasingly efficient resulting in continued power loss reductions while increasing our solution densities.

 

Infineon digital controllers bring unprecedented flexibility and adaptability as well as precise control, telemetry and protection features. As a leader in this AI power delivery market, Infineon offers a broad range of controllers and OptiMOS™ power stages that can support all known AI hardware platforms and their demanding current levels. Infineon enables designers to create state-of-the-art power solutions with highest efficiency and power density for today’s high power AI applications as well as future needs.

 

The machine learning cluster at the University of Waikato is work deep learning during a range of how. The companies have desegregated many state-of-the-art deep learning approaches into the well-known machine learning software made by the cluster. The aim is to create deep learning doable without requiring the user to jot down programs. The cluster is also staring at the employment of transfer learning in many deep learning applications. Transfer learning is that the plan of adding coaching data that’s specific to a definite task to a pre-existing learned model. For example, if the task is characteristic an object in a picture there are in public available models trained on large libraries of images. In the New Zealand context they will be tailored to be used in agriculture, like estimating yield in grape production.

 

The idea is that the pre-existing network has already learned to extract useful options from general pictures, like edges, textures and shapes, etc. By further coaching of the model exploitation pictures of grapes and surrounding vegetation, it becomes doable to be told higher-level options that area unit specific to grapes.

 

It is then comparatively simple to count individual grapes and estimate yield. The advantage of this approach is that the grape information set are often significantly smaller than the one accustomed establish the pre-existing model.

Leave a Reply

Your email address will not be published. Required fields are marked *