Drone Landing Identification an Intel AI Reference Kit Lab

This lab will walk you through one of Intel's AI Reference Kits to develop an optimized semantic segmentation solution based on the Visual Geometry Group (VGG)-UNET architecture, aimed at assisting drones in safely landing by identifying and segmenting paved areas. The proposed system utilizes Intel® oneDNN optimized TensorFlow to accelerate the training and inference performance of drones equipped with Intel hardware. Additionally, Intel® Neural Compressor is applied to compress the trained segmentation model to further enhance inference speed. Explore the Developer Catalog for information on various use cases.

What you will see:

• Training and inferring with complex neural networks like the VGG-UNET model can be computationally demanding. In this reference kit, TensorFlow is optimized using Intel oneDNN to enhance performance on Intel hardware during both training and inference stages.
• Intel® Neural Compressor library further boosts inference efficiency. This project utilizes it to convert the trained FP32 VGG-UNET model into an INT8 CRNN model through post-training quantization, reducing model size and enhancing inference speed.

• To understand the benefits of Intel's OpenDNN and Neural Compressor Library
• Leverage Intel's Xeon 5th Generation processors to train, tune, and inference the model.

• Intel 5th Generation Processors on Dell PowerEdge R660 Servers
• Intel oneDNN library
• Intel® Neural Compressor library
• OpenShift AI