📢 Announcement: MemryX SDK 2.0 is now released!
Please update to the latest SDK version before proceeding with any of the examples.

Welcome to MemryX eXamples, a collection of end-to-end AI applications and tasks powered by MemryX hardware and software solutions. Whether you're performing real-time video inference, exploring fun AI projects, or generating text, these examples provide practical, hands-on use cases to help you fully leverage MemryX technology. For detailed guides and tutorials, visit the MemryX Developer Hub.

To ensure a smooth experience with MemryX solutions, follow these steps before diving into the examples:

• Explore the Developer Hub: Your gateway to comprehensive documentation for MemryX hardware and software.
• Install the MemryX SDK: Set up the essential tools and drivers to begin using MemryX accelerators.
• Check out our Tutorials: Step-by-step instructions for various use cases and end-to-end applications.
• Explore the Model Explorer: A great starting point for discovering models compiled and tested on MemryX accelerators.

Before working with the examples, ensure your system is correctly set up by installing the MemryX SDK. Follow the detailed instructions here: MemryX SDK Get Started Guide.

Clone this repository plus any linked submodules with:

git clone --recursive https://github.com/memryx/memryx_examples.git

Note: Applications marked with 📝 have tutorials available. Clicking on the icon will take you directly to the tutorial page.

Leverage MemryX accelerators for real-time video processing tasks. These applications demonstrate how to run models efficiently on live video streams.

Application	Description	Models	Code	OS	Preview
Depth Estimation using MiDaS 📝	Estimate depth from a video stream	MiDaS
Object Blurring using YOLOv8	Protect individuals privacy in a video stream	YOLOv8
Object Tracking using YOLOv8	Object tracking in a video stream	YOLOv8
Object Detection using YOLOv7 📝	Detect objects in real time	YOLOv7 (Tiny)
Object Detection using CenterNet 📝	Detect objects in real time	CenterNet
Object Detection using YoloX	Detect objects in real time	YoloX (Medium)
Vehicle Detection	Detect vehicle in real time	Vehicle-Detection-0200
Segmentation using YOLOv8	Perform instant segmentation on video in real time	YOLOv8 Nano Segmentation
Pose Estimation using YOLOv8 📝	Estimate human pose from video	YOLOv8 (Medium)
3D Point Cloud from Depth Estimation	Generate real-time point clouds from depth data	MiDaS
Wireframe detection Using M-LSD and QT	Perform Line segment detection in real time	M-LSD (Large)
Face Detection & Emotion Classification 📝	Detect faces and classify emotions	Multiple Models
Person Tracking using YOLOv7	Track unique people across video frames	YOLOv7 (Tiny)
Face Landmarks Detection	Detect presence of a face and its landmarks	BlazeFace and FaceMesh
Mediapipe Hand Landmarks	Detect hands and draw a "skeleton" of landmarks	PalmDet and HandPose
Interactive Realtime Multi-Face Recognition	Interactive app for face recognition	Multiple Models
Intrusion Detection	Detect any intruding object in a ROI	Yolov8 and ByteTrack

Explore models performing inference on static images and data. These examples demonstrate how to leverage the MXA to process large amounts of data.

Application	Description	Models	Code	OS	Preview
Satellite Object Detection with Oriented Boxes	Detect oriented bounding boxes on satellite images	YoloV8m-OBB
Face Detection + Recognition	Perform face detection + recognition	YoloV8n-Face + FaceNet

Explore how the MemryX accelerators can be used for open-vocabulary tasks. The examples below demonstrate how to run models using MemryX hardware.

Application	Description	Models	Code	OS	Preview
Open-Vocab Seg with YoloE	open-vocabulary, zero-shot object detection and segmentation	yoloe-v8s-seg
Zero-Shot Classify — OpenAI CLIP	open-vocabulary, zero-shot object classification	ClipResNet50

Maximize performance by running multiple video streams concurrently on MemryX accelerators.

Application	Description	Models	Code	OS	Preview
Multi-Stream Object Detection using YOLOv8S 📝	Detect objects across multiple streams	YOLOv8 (Small)
Multi-Stream Object Detection using YOLOv7Tiny 📝	Detect objects across multiple streams	YOLOv7 (Tiny)

Run multiple models simultaneously on separate DFPs within a single application.

Application	Description	Models	Code	OS	Preview
Cartoonizer and Pose (Side-by-Side)	Real-time cartoon effects and pose detection simultaneously	Facial-Cartoonizer & YOLOv8s-pose
Cartoonizer and Pose (Selectable Overlays)	Real-time overlay of cartoon and pose estimation	Facial-Cartoonizer & YOLOv8s-pose

Configurable many-stream video applications with USB, RTSP, and video file support, using optimized C++ pre/post-processing.

Application	Description	Models	Code	OS	Preview
Object Detection with YOLOv8	High performance, configurable C++ object detection	YOLOv8 (Nano)
PCB Defect Detection	Detect defects on printed circuit boards in real time across multiple video streams	Retrained YOLOv8 (Small)

Explore interactive and engaging AI-powered applications in our fun projects section.

Application	Description	Models	Code	OS	Preview
Chrome Dino Game 📝	Control the Chrome Dino Game using palm detection	Palm Detection
Deep Reinforcement Learning with Mario	Play Mario with a Reinforcement Learning Agent	Custom
Aimbot	Automatic aim and click for Windows games	YOLOv7 (Tiny)
Repcounting Web-application	Workout repcounting web-application	YOLOv8 Pose (medium)
Facial Cartoonizer	Instantly cartoonize videos in real time	Facial-Cartoonizer
Virtual Painter	Virtually Paint using Hand Landmarks in real-time	Palm Detection & Hand Landmark
ASL Alphabet to Text	Use the American Sign Language alphabet to spell words	Palm Detection & Hand Landmark

Measure and evaluate the accuracy of various models using MemryX hardware.

Task	Description	Models	Code	OS
Classification Accuracy 📝	Calculate accuracy for classification models	ResNet50
Object Detection Accuracy 📝	Calculate accuracy for object detection models	YOLOv8 (Medium)
Keras Classifiers Accuracy 📝	Calculate Keras classifiers accuracy on the MXA	Keras applications

Audio Inference 🔊

Explore how the MemryX accelerators can be used for audio/speech processing tasks. The examples below demonstrate how to run models using MemryX hardware.

Task	Description	Models	Code	OS
Audio Denoising using UNet	Remove noise from speech audio using UNet	UNet
Audio Classification using YAMNet	Identify audio categories using YAMNet	YAMNet
Audio Classification Web App	Classify audio using YAMNet in a web app	YAMNet
Speech Emotion Recognition Web App	Detect emotion from speech (web app)	Light-SERNet

Useful Links

• Developer Hub — Comprehensive documentation for MemryX hardware and software.
• DevHub Get Started — Guide to set up MemryX software and hardware.
• Tutorials — Step-by-step instructions for various use cases and applications.
• FAQ — Frequently asked questions.
• Troubleshooting Guide — Solutions to common issues.

Contribution Guidelines

We welcome contributions! If you'd like to contribute to this repository or examples, please refer to our contribution guidelines. Feel free to submit pull requests, suggest improvements, or ask questions in the issues section.

Frequently Asked Questions (FAQ)

1. How do I install the MemryX SDK?

Refer to the SDK Installation Guide for a detailed step-by-step guide on setting up the MemryX SDK.

2. What do I do if an example isn't working?

Make sure you’ve followed all setup steps. You can also check the Troubleshooting Guide for more help, or open an issue in the repository.

3. Can I contribute to this repository?

Yes! We welcome contributions. Please refer to our contribution guidelines for more information on how to contribute.

Happy coding! 😊
The MemryX Team