>5s Each Image - Processing Performance - How Improve?

Hello! Can anyone please explain me why my system takes a lot of seconds to process 1 image low resolution in Deepstack?

Lots of people telling me it tooks miliseconds, but mine is taking seconds…don’t know why!

DeepStack: Version 2021.01.0
/v1/vision/detection

My host is a Dell Optiplex 3070 with i5 9500t 6core 24Gb DDR4 Ram, the proxmox Win2019 VM has 4VCPU and 8Gb RAM. No GPU.

Idle: 15 to 20%, W19 with Blueiris 8 cam streams, AiTool, Deepstack.

Images for Deepstack ~640x480

Running in Low Mode: 2 to 3 seconds (deepstack --VISION-DETECTION True --MODE Low --PORT 83)
Running in Standard Mode: 5 to 6 seconds (deepstack --VISION-DETECTION True --PORT 83)

You’re using the CPU which uses a lot of CPU power. Get a cheap graphics card like a GTX 10XX I got a 1060 3gb ram refurbished on amazon for like $180. Graphic cards even older ones are in short supply due to the shortage. I can process images on the GPU 40 to 80ms on high 1080p image only using like 3% of the GPU. I use cross zone duplicate camera’s hidden on blue iris so any objects that cross from zone A>B,B>A and have it save a image every 0.5 seconds no video just images for 4 seconds trigger so that’s about 8 images everyone once in awhile if someone triggers the camera or a shadow or lights at night. Then normal camera’s to record motion no alerts. Then I have AI tools send a image to telegram group and trigger blue iris hidden camera to alert me if there is humans or car in masked area using https://github.com/VorlonCD/bi-aidetection the masks work in that version.