Is there a thread for v17?

[pdawg17]

I have some questions :)

[hawkeye217]

No. 0.17 is still in very early development. We are not yet ready to accept feedback or bug reports on anything related to it. When we are ready, we'll post a discussion.

[xavan83]

According to the metrics page, the current dev branch reduces inference time by 25%, dropping from 12 ms to 9 ms on my i3-12100 with iGPU using the OpenVINO detector.

[NickM-27]

Glad to hear it!

[xavan83]

But same model on onnx gives 40msec (no matter why, I don't use it, it is just for testing) openvino is fine

[NickM-27]

Onnx doesn't use the GPU for Intel

[pdawg17]

Onnx doesn't use the GPU for Intel

So you're saying that with my 10100T, this won't use the GPU?

detectors:
ov:
type: openvino
device: GPU

model:
model_type: yolonas
width: 320
height: 320
input_tensor: nchw
input_pixel_format: bgr
path: /config/model_cache/yolo_nas_s.onnx
labelmap_path: /labelmap/coco-80.txt

If not, what's the best alternative for v17?

[hawkeye217]

Onnx doesn't use the GPU for Intel

So you're saying that with my 10100T, this won't use the GPU?

detectors:

ov:

type: openvino

device: GPU

model:

model_type: yolonas

width: 320

height: 320

input_tensor: nchw

input_pixel_format: bgr

path: /config/model_cache/yolo_nas_s.onnx

labelmap_path: /labelmap/coco-80.txt

If not, what's the best alternative for v17?

No, this will use your iGPU because you are using type: openvino. That is different than type: onnx.

[comatose-tortoise]

How do i find the docker images for the latest dev builds? How do i know which ones are 0.17?

[NickM-27]

They're in the GitHub sidebar, you know they're 0.17 based on the commit and which branch the commit belongs to (dev or master)

In general, 0.17 is actively being worked on and not considered stable so use at your own risk