yolov8测试脚本

以下是yolov8测试脚本的示例代码：

import cv2
import numpy as np
import onnxruntime as ort

# 加载模型
sess = ort.InferenceSession('models/yolov8s.onnx')

# 加载类别名称
with open('yolov8_onnx/coco.names', 'r') as f:
    class_names = [cname.strip() for cname in f.readlines()]

# 定义输入和输出节点名称
input_name = sess.get_inputs()[0].name
output_names = [sess.get_outputs()[i].name for i in range(len(sess.get_outputs()))]

# 定义预处理函数
def preprocess(img):
    # 缩放图像
    img = cv2.resize(img, (416, 416))
    # 转换颜色空间
    img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
    # 归一化像素值
    img = img.astype(np.float32) / 255.0
    # 转换维度顺序
    img = np.transpose(img, [2, 0, 1])
    # 添加批次维度
    img = np.expand_dims(img, axis=0)
    return img

# 定义后处理函数
def postprocess(outputs, conf_thresh=0.5, nms_thresh=0.5):
    # 解析输出
    boxes, scores, classes = [], [], []
    for output in outputs:
        output = np.squeeze(output)
        num_classes = output.shape[1] - 5
        for i in range(output.shape[0]):
            box = output[i, :4]
            score = output[i, 4]
            class_idx = np.argmax(output[i, 5:])
            class_score = output[i,5 + class_idx]
            if score * class_score > conf_thresh:
                x1, y1, x2, y2 = box
                boxes.append([x1, y1, x2, y2])
                scores.append(score * class_score)
                classes.append(class_idx)

    # 非极大值抑制
    keep = cv2.dnn.NMSBoxes(boxes, scores, conf_thresh, nms_thresh)

    # 构建检测结果
    results = []
    if len(keep) > 0:
        for idx in keep.flatten():
            x1, y1, x2, y2 = boxes[idx]
            score = scores[idx]
            class_idx = classes[idx]
            class_name = class_names[class_idx]
            results.append({'box': [x1, y1, x2, y2], 'score': score, 'class': class_name})

    return results

# 加载图像
img = cv2.imread('yolov8_onnx/dog.jpg')

# 预处理图像
img = preprocess(img)

# 运行模型
outputs = sess.run(output_names, {input_name: img})

# 后处理输出
results = postprocess(outputs)

# 显示结果
for result in results:
    x1, y1, x2, y2 = result['box']
    score = result['score']
    class_name = result['class']
    cv2.rectangle(img, (x1, y1), (x2, y2), (0, 255, 0), 2)
    cv2.putText(img, f'{class_name} {score:.2f}', (x1, y1 - 10), cv2.FONT_HERSHEY_SIMPLEX, 0.5, (0, 255, 0), 2)

cv2.imshow('result', img)
cv2.waitKey(0)
cv2.destroyAllWindows()

该脚本可以加载yolov8的onnx模型，并对一张图片进行推理，输出检测结果。在预处理函数中，将图像缩放到416x416大小，并归一化像素值；在后处理函数中，解析模型输出，进行非极大值抑制，并构建检测结果。最后，将检测结果绘制在原图上并显示出来。