Yes, our system can do video and real-time facial recognition. The frame rate is normally around 30fps. But the frame rate will be changed depending on the video.
For your first question, we hope the accuracy of the facial recognition part can be achieved at least 90% so that it can be used. But our goal is to improve the accuracy to 95-98% so that the system can be more accurate.
For your second question, we think it is feasible to achieve 90% accuracy. In fact, we have reached 90% accuracy for recognizing 3 people simultaneously. Now we are trying to increase the number of people for recognition.
SINGH, Dilsher
January 19, 2022 2:31 pm
Could you describe in detail about the main technology used?
Hello Signh,
Thanks for your comments. For hardware, we will use the Ultra96 V2 embedded board to run the program. For the facial recognition part, we use the deep learning algorithms like CNN to recognize human faces as Ultra 96 V2 provides a DPU which can accelerate the process. We will also combine both camera images for image registration, so the screen can show a person’s identity and body temperature simultaneously. we prefer using the FPGA in the board to run this function.
Thanks for your question, one of the main technology is the FPGA board to combine the two cameras.
Our project achieves facial recognition and temperature taking. However, those two cameras are different pixels sizes so we use pixel matching to implement the higher matching rate.
Also, the thermal camera has some thermal energy loss through taking. We readjust the way of taking temperature so it can fit the true temperature no matter measure which part of the body.
Can your system process video in real-time as of today? If yes, what is the frame rate?
Hello Keith, We can do real time facial recognition with ~15fps.
Thanks for your questions.
Yes, our system can do video and real-time facial recognition. The frame rate is normally around 30fps. But the frame rate will be changed depending on the video.
It will be nice to have the framerates under different situations in your final presentation.
From the progress part, I recognise the accuracy is around 70% and you mention that it can be improved by using more samples/data.
I have two simple question on this area
Thanks very much
Thanks for your questions.
For your first question, we hope the accuracy of the facial recognition part can be achieved at least 90% so that it can be used. But our goal is to improve the accuracy to 95-98% so that the system can be more accurate.
For your second question, we think it is feasible to achieve 90% accuracy. In fact, we have reached 90% accuracy for recognizing 3 people simultaneously. Now we are trying to increase the number of people for recognition.
Could you describe in detail about the main technology used?
Hello Signh,
Thanks for your comments. For hardware, we will use the Ultra96 V2 embedded board to run the program. For the facial recognition part, we use the deep learning algorithms like CNN to recognize human faces as Ultra 96 V2 provides a DPU which can accelerate the process. We will also combine both camera images for image registration, so the screen can show a person’s identity and body temperature simultaneously. we prefer using the FPGA in the board to run this function.
Thanks for your question, one of the main technology is the FPGA board to combine the two cameras.
Our project achieves facial recognition and temperature taking. However, those two cameras are different pixels sizes so we use pixel matching to implement the higher matching rate.
Also, the thermal camera has some thermal energy loss through taking. We readjust the way of taking temperature so it can fit the true temperature no matter measure which part of the body.