Skip to content

Latest commit

 

History

History
30 lines (22 loc) · 1.25 KB

File metadata and controls

30 lines (22 loc) · 1.25 KB

Deep-learning with DJL and Apache MXNet

Neural networks with numerous layers of nodes allow for more complex, rich and deeper processing and understanding. This example detects objects within an image. It uses a pre-trained model and the Deep Java Library backed by the Apache MXNet engine.

MXNet.groovy

Groovy code examples can be found in the DeepLearningMxnet subproject. If you have opened the repo in IntelliJ (or your favourite IDE) you should be able to execute the examples directly in the IDE.

Requirements: The code has been tested on JDK8, JDK11 and JDK17.