search

marcoslucianops / DeepStream-Yolo

NVIDIA DeepStream SDK 7.0 / 6.4 / 6.3 / 6.2 / 6.1.1 / 6.1 / 6.0.1 / 6.0 / 5.1 implementation for YOLO models
MIT License
1.48k stars 357 forks source link

about reducing memory when running #25

Closed hejun888 closed 3 years ago

hejun888 commented 3 years ago

hi,why the whole process takes up so much memory, and is there any way to reduce the memory

marcoslucianops commented 3 years ago

Are you using Jetson platform?

BarcaBear commented 3 years ago

I'm seeing a similar issue on an Xavier NX. The detection is much better then YoloV3 but it's slower. Around 20fps vrs 30fps for YoloV3. YoloV4-CSP seems to be the fastest. Tried fp16/32 and int8. But don't notice much difference.

Would be great to achieve 25fps with same level of detection.

Any tweaks you can think of ?

hejun888 commented 3 years ago

yes i use jetson platform

marcoslucianops commented 3 years ago

Idk how to reduce deepstream memory usage, but you can reduce system memory usage by using lxde (see my video: https://youtu.be/Z3K43Q34sIs). Btw, the system will have more memory to use during deepstream/post process.