Object detetction, tracking and counting using the Pi

[ThohtDelta]

Hello,
I've been working on a project which counts people. When I run the codes on a pre-captured video feed, the system performs in real-time. But, when I try to implement the code by using a live feed from the pi-camera, the system lags and results in less accurate results.
Is this the due to the computational potentiality of the pi?

Can someone help me solve this issue?
Thanks in advance!

[B.Goode]

Can someone help me solve this issue?

On the basis of the information provided that seems unlikely!



Maybe provide a few hard facts.


What model of RPi?

What Operating System?

What libraries or tools are you using to capture and analyse the video feed?

[ThohtDelta]

Thanks for the reply!

Well, I use
Rpi model : B+
OS : Raspbian Stretch
Camera : Pi Camera Module v2
Libraries : picamera, opencv, imutils

Let me know if you need anymore info on the same.

[Joel_Mckay]

Try YOLO:
https://www.raspberrypi.org/forums/view ... p?t=219601

[B.Goode]

Well, I use
Rpi model : B+

Does that mean this -
https://www.raspberrypi.org/products/r ... el-b-plus/


Your hunch may be correct. Maybe this model, superceded over 3 years ago, does not have the compute power needed for your non-trivial application.

Have you considered replacing it with the current top of the range and most powerful ever model: https://www.raspberrypi.org/products/ra ... el-b-plus/

[ThohtDelta]

Well, I've been using Rpi 3 B+ model only.