This user evaluation report of MYIR’s MYD-JX8MPQ development board is provided by “Liu Xing Ke Ji” from EEWORLD. The MYD-JX8MPQ development board consists of a high-performance CPU Module MYC-JX8MPQ and a base board to provide a complete evaluation platform for NXP’s i.MX 8M Plus application processors which features quad Arm Cortex-A53 cores running at up to 1.8 GHz and an 800 MHz Cortex-M7 real-time co-processor, with an integrated Neural Processing Unit (NPU) operating at up to 2.3 TOPS, 800 MHz audio DSP, 1080p VPU, dual ISPs and 3D/2D GPU to focus on machine learning and vision, Artificial Intelligence (AI), advanced multimedia and industrial automation fields. This report is for face recognition based on QT+OpenCV on MYD-JX8MPQ board.
MYD-JX8MPQ Development Board
The tested open source project is based on QT+OpenCV face recognition punch project. The used open source code was originally running under WIN, but was slightly modified to run on MYIR’s MYD-JX8MPQ for testing purpose.
The tested project is actually divided into two projects. One project is used as an administrator control function to add face information. At the same time, you can also query the clocking record, and send notifications to the slave machine.
OpenCV face detection classifier is mainly needed to be used for face recognition.
Better recolonization can only be implemented after getting the training model. Here a brief introduction was given for the training.
Through the above code, image acquisition is completed.
Training with the code above, using Python. Therefore, the system environment needs to be configured.
In this file, take the images that we've collected, put them in, and create a new folder.
The next step is to add at.txt to our file.
After training, we have the training documents we need.
This is the punch interface when we punch in. Load the trained stuff and start the timer to get the camera signal. Finally consistent with the database is considered to be successful.
The training part above, another project provided is actually all completed.
This is from our WIN interface, and the round box is the position where the image captured by our camera will be displayed.
We need to replace all the libraries under Ubuntu so they can be compiled and then copied to the development board to run. As follows:
The message “database not open” is presented. We are based on the database, so it is still more embarrassing, later if you can try to compile all and update it. Let's test it now and see what happens.
The hardware used adds a camera.
This is what it looks like when you turn on the camera.
This GIF shows us face detection.
Since there is no database, I can only print some information. When the two numbers are equal, the next step is to judge to punch in. Let's show the effect on the computer side.
|
