How to improve the processing

nicholaskajoh / ivy

Video-based object counting software.

MIT License

428 stars 170 forks source link

Closed whitetiger21022014 closed 4 years ago

whitetiger21022014 commented 4 years ago

Thank for your code But it's so slow I see the document as below "Real-time and video processing object detection using Tensorflow, OpenCV and Docker. https://towardsdatascience.com/real-time-and-video-processing-object-detection-using-tensorflow-opencv-and-docker-2be1694726e5 "

Do you think if we use Docker as that document the speed will be improved or not? Thank you

nicholaskajoh commented 4 years ago

No, Docker won't improve the processing speed.

To get better performance:

Use a GPU. You can currently do this with either the tfoda (TensorFlow) or detectron2 (PyTorch) detector.
Train a light-weight model. The model used in the demo is quite large and most of it is not being utilized.
Use input video of lower framerate/resolution. This might negatively impact accuracy though.