Object Tracking is a combination of detecting object and assigning unique identity. In each consecutive frame the object will maintain its identity. Multi-object tracking (MOT) aims at tracking objects of a single class or multiple classes. Here an object can be like persion, vehicle, luggage etc.
- Qualcomm® Robotics RB5 Introduction Guide:
- The Qualcomm® Neural Processing SDK provides tools for model conversion(onnx to dlc), model quantization and execution. Refer to the installation steps given in the detailed documentation from the SDK.
- ADB to access the device using command prompt and push/pull the files from device.
- An Ubuntu 20.04 PC
- Qualcomm Robotics RB5 Development kit: https://developer.qualcomm.com/qualcomm-robotics-rb5-kit
- A USB camera
- A display monitor
The application implemented People Tracking using ByteTrack model:
Person tracking
This solution will track persons in a frame by assigning unique tracker ID
demo.mp4
Ref: https://github.com/ifzhang/ByteTrack
This application supports object detection use cases with YoloX model using Qualcomm® Neural Processing SDK.
Refer Design.md to understand the implementation details and steps to integrate a model into the application. Design Details
There are prerequisite and some packages need to be installed to run the application. Please refer Install.md to prepare the setup on development platform. Setup Device
Need to prepare model for the solution. Refer ModelPreperation.md for model generation. Model Preperation
After setup is done. The application can be compiled and executed. Follow the instructions given in Build.md to execute the application on development platform. Compile and deploy