Wahyu Setianto

Hello, this is the issue from my repo [Hyuto/yolo-nas-onnx](https://github.com/Hyuto/yolo-nas-onnx) to inference YOLO-NAS ONNX model. I found the main problem is the preprocessing steps @Uaena-Wang custom trained model isn't the same...

Hello, I think live detection performance depend to the device component (CPU/GPU). You may experience slow detection on low end android devices, but on high end it's pretty fast. I...

Well you can't compare the web and app version, even using the same model will not close the gap of performance. App version is low level so performance is peak...

Hello, well it's definitely possible to stream video through webcam, firstly using `opencvjs` to take every frame from webcam and then do inference just like in the code for every...

Hello, Exporting YOLOv8 model to TFJS can be done using `export` method, as you can see on the documentation [Export Usage Examples](https://docs.ultralytics.com/modes/export/#usage-examples) with `format="tfjs"`. **Example** ```python from ultralytics import YOLO...

Yes, the input shape of all TF model should be with [batch, height, width, channels] format, I don't know why your model is using [batch, channels, height, width] format which...

Well that's normal because YOLOv8 ONNX model input shape is [b, c, h, w] format, and you just convert it raw. I suggest to `transpose` the input image tersor before...

Hello, I think there is no problem on build. I suspect the main issue is you try to serve build dist on "." base level since this project is serve...

Hello, I doubt we can do a lot of these type in JS, but if you want to get the final mask it's on https://github.com/Hyuto/yolov8-seg-tfjs/blob/4d5c0c216ecf0a9eb8313ef4b9a18de831d49068/src/utils/detect.js#L144 `mask` variable contains boolean typed...

Hello, I think that's a good idea to solve it on the frontend but the computation of every device isn't the same so you must take attention of that. For...