search

ultralytics / yolov3

YOLOv3 in PyTorch > ONNX > CoreML > TFLite
https://docs.ultralytics.com
GNU Affero General Public License v3.0
10.25k stars 3.45k forks source link

Real Time Human Face Detection using IP Camera #564

Closed MrFahad closed 5 years ago

MrFahad commented 5 years ago

I am using yolov3 model for the human face detection. For webcam it works fine but when I gave any video or any IP Camera address as rtsp it works very slow.

For video it reads all the video till end but in rtsp when I want to get real time streaming it starts live streaming frame per seconds and after few frame it stops and close. And one thing read video very slowly.

Kindly guide me. Live-Streaming

glenn-jocher commented 5 years ago

This depends entirely on your hardware. Our iOS app runs YOLOv3-SPP-320 at 30 FPS on iPhone 11 for example.

MrFahad commented 5 years ago

Kindly suggest me hardware, as right now I have 4 IP cameras . I tried all over the z800 workstation with 32gb ram and Nvidia 8 GB gpu 1070.

Still real time streaming is slow as on my core i3 laptop.

Kindly guide me.

Get Outlook for Android

On Fri, Oct 25, 2019 at 10:15 PM +0500, "Glenn Jocher" notifications@github.com wrote:

Closed #564.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or unsubscribe.

frankcarey commented 5 years ago

Ensure you are running on GPU. You can do this via your python REPL.

$ python
>> import torch
>> torch.cuda.device_count()

you should get back 1

See https://stackoverflow.com/questions/48152674/how-to-check-if-pytorch-is-using-the-gpu

glenn-jocher commented 5 years ago

This repo will use any available GPU(s) detected.