How to use yolov5x-seg.onnx?
winktool opened this issue · 10 comments
The model link is : https://github.com/ultralytics/yolov5/releases/download/v7.0/yolov5x-seg.pt
How to change these codes for getting masks?
const dims = [1, 3, height, width];
const imageTensor = await imageDataToTensor(image, dims);
const feeds = {};
feeds[session.inputNames[0]] = imageTensor;
const results = await session.run(feeds);
const output0 = results[session.outputNames[0]];
const boxes = [];
for (let r = 0; r < output0.data.length; r += output0.dims[2]) {
const data = output0.data.slice(r, r + output0.dims[2]); // get rows
console.log("data", output0.data);
const scores = data.slice(5); // classes probability scores
const confidence = data[4]; // detection confidence
const classId = scores.indexOf(Math.max(...scores)); // class id of maximum probability scores
const maxClassProb = scores[classId]; // maximum probability scores
if (confidence >= confidenceThreshold && maxClassProb >= classThreshold) {
const [x, y, w, h] = data.slice(0, 4);
boxes.push({
name: labels[classId],
score: (maxClassProb * 100).toFixed(1),
classId: classId,
probability: maxClassProb,
confidence: confidence,
bounding: [x - 0.5 * w, y - 0.5 * h, w, h],
});
}
}
Like this: https://github.com/UNeedCryDear/yolov5-seg-opencv-dnn-cpp/blob/master/yolo_seg.cpp#L115
Hey,bro,i found these,maybe can help us.
ultralytics/yolov5#9746
Hello, sorry but I can't help you on this. Since I don't have any experience on object segmentation case, I still don't know how the model really work. Object segmentation is a new and exciting project, I'll surely do it in the future project but not in short time.
Yes, I'm looking forward to your new progress.
Maybe this page can help us:
https://blog.csdn.net/qq_41043389/article/details/127754384
I've create new repo on deploy YOLOv5-seg on onnxruntime but in python. You can take a look here https://github.com/Hyuto/yolov5-seg-onnxruntime
You're a really good guy!
but do you know how to use them with nodejs.
python version link:https://github.com/ultralytics/yolov5/blob/master/segment/predict.py
Hello just a little bit update here, I've implemented it using onnxruntime-web although the performance is bad and can't do live detection (video), but hey IT WORKS 😁
https://github.com/Hyuto/yolov5-seg-onnxruntime-web
If any issue on yolov5-seg you can put it there. I'm closing this issue