playqt - Computer Vision Real Time Object Detection HD

https://sourceforge.net/projects/playqt/ playqt is a media player with a built in YOLO AI model for Object Counting. Real time traffic analysis or alarm applications. Data is output in format compatible with Excel. Stable production quality open source code. The core system is a gui built around the famous command line program ffplay, and may be invoked from the command line using the same options. Developers may use this platform for building their own applications by adding filters to the system. The source code may be easily compiled using standard tools. Qt6, MSVC 2019, NVIDA Cuda Development library are required. MODEL CONFIGURATION The program must be configured after installation to connect to the AI model. A sample model is provided at https://sourceforge.net/projects/playqt/files/. Download and unzip the model.zip file in your installation directory. By default this will be %HOMEPATH%/playqt. Once the model has been unzipped, launch the program and use the menu bar at the top to go to Tools--Filters. On the right hand top side of the filter screen are the available filters, double click on darknet to activate the AI model screen. There are three files that comprise the model Names, Weight and Config. The three directory boxes are used to point to those files. The three dot buttons on the right hand of the directory fields may be used to navigate to the files. The model resolution can be adjusted in even increments of 32. Higher resolution will improve accuracy at the expense of longer processing times / lower frames per second. The threshold is the model confidence detection threshold. Lower thresholds will increase the number of objects detected at the expense of false positives. To learn more about darknet, visit https://github.com/AlexeyAB/darknet for an excellent description of the framework and resources for developing customized models. CAMERA CONFIGURATION Full control of ONVIF compatible cameras is available from the Cameras tab on the right side of the main screen. The config sub tab can be used to implement automatic discovery and launch of cameras. If all cameras are configured to use the same user name and password, the common boxes may be filled for the info to be used during discovery. If the Auto Discovery box is checked, playqt will find the cameras on startup. The Auto Load Camera field nay specify a camera to start automatically at launch. If the host computer has more than one network interface, it is possible to specify which network to use during the discovery. This feature is especially useful for isolating cameras on a subnet, a good practice that can improve performance and security. FOR DEVELOPERS playqt mar be easily compiled from source. Qt6, MSVC 2019 and the NVIDA GPU Computing Toolkit version 11.2 are required. The contrib zip file contains the development and run time binaries needed to compile on the windows platform. The source code is configured to work with Qt