Starter Applications¶
The alwaysAI platform provides applications that cover basic usage of all the major APIs. This page describes what the starter apps accomplish in more detail; you can read how to set up sample Projects and starter applications on the Using Starter Applications page.
The best way to confirm that your system is set up properly to install and run alwaysAI applications is to try running the Hello World application, which performs object detection on a set of images. Next, you can explore the edgeIQ API with the following apps.
Image Classification¶
Image classifier¶
This app performs image classification on a set of images, and is a great starting point for a use case requiring batch processing on many images.
Age and gender classifier¶
This app performs two classifications on each image of a set of faces. This is a useful reference design for any use case that requires running multiple inferences on an image or video frame using different models.
Object Detection¶
Object detector¶
This app performs object detection on a set of images.
Real-time object detector¶
This app performs object detection on a real-time video stream.
Real-time face detector¶
This app performs object detection using a model trained only to detect faces, on a real-time video stream.
NVIDIA real-time object detector (for use with the Jetson Nano)¶
This app performs object detection on a real-time video stream. It uses the CUDA engine for inferencing.
Pose Estimation¶
Real-time pose estimator¶
This app performs pose estimation on a real-time video stream.
Semantic Segmentation¶
Semantic segmentation with the VOC dataset¶
This app performs semantic segmantion using a model trained on the VOC dataset, on a set of images.
Semantic segmentation with the Cityscapes dataset¶
This app performs semantic segmentation using a model trained on the Cityscapes dataset, on a set of images.
NVIDIA autonomous vehicle semantic segmentation (for use with the Jetson Nano)¶
This app performs semantic segmantion on a video clip of a driving scene. It uses the CUDA engine for inferencing.
Object Counting and Tracking¶
Object counter¶
This app performs object detection on a real-time video stream, and counts the number of each object visible in the frame.
Face counter¶
This app performs [object detection](application_development/core_computer_vision_services:Object Detection) on a real-time video stream using a model trained to detect only faces, and uniquely tracks each face in the frame using the centroid tracker.
Object detector and tracker¶
This app periodically performs object detection on real-time video, and performs correlation tracking between each detection. This app is a useful reference for a use case where detections are very slow but object location is needed at a more frequent update rate.
Run the Starter Apps¶
Follow these instructions to learn how to install and run the alwaysAI starter apps.