基于yolov5获取图片坐标

Sanyueshui_Go

于 2024-07-02 17:31:16 发布

阅读量203

点赞数 4

文章标签： YOLO 深度学习人工智能

本文链接：https://blog.csdn.net/m0_63573177/article/details/140132267

版权

import torch


model = torch.hub.load('.', 'custom', path='./runs/train/exp17/weights/best.pt',source="local") # 加载模型

model.eval()

img = './datasets/images/train/img_16.png' # 需要被推理的图片


results = model(img)

detections = results.xyxy[0]  

SLIDER_CLASS_INDEX = 0


slider_boxes = []
for *box, conf, cls in detections:
    if cls == SLIDER_CLASS_INDEX:  
        x1, y1, x2, y2 = map(int, box)
        slider_boxes.append((x1, y1, x2, y2))


for i, box in enumerate(slider_boxes):
    print(f'Slider {i+1}: {box}')

运行并输出

YOLOv5  2021-10-12 torch 1.9.0+cu111 CUDA:0 (NVIDIA GeForce RTX 3090, 24575.375MB)

Fusing layers... 
Model Summary: 213 layers, 7012822 parameters, 0 gradients, 15.8 GFLOPs
Adding AutoShape... 
Slider 1: (225, 44, 284, 93) # 检测出的坐标