yolov11 epoch100轮训练笔记5 kaggle comet

njsgcs

已于 2025-05-05 13:37:25 修改

阅读量281

点赞数 7

分类专栏： yolo笔记文章标签： YOLO 笔记人工智能

于 2025-05-05 13:24:21 首次发布

本文链接：https://blog.csdn.net/njsgcs/article/details/147710653

版权

yolo笔记专栏收录该内容

8 篇文章

订阅专栏

Football Players Detection using YOLOV11 | Kaggle

!pip install comet_ml


import comet_ml

comet_ml.login(project_name="c")

Comet - Build Better Models Faster

yolov11训练

100轮一眨眼训练完了

然而comet接不到yolo的sdk

优秀

训练17轮map就0.99了

v5训练100轮才0.6

有了，就是训练轮次不够

效果和11比差远了

map在50轮后才稳定上升

有边框信息了

本地detect读取边界框信息

w torch==2.5.1 torchvision --index-url https://download.pytorch.org/whl/cu124
w ultralytics

~~windows下torch能要我命睡个午觉先~~

from ultralytics import YOLO
import numpy as np
# 加载模型
model = YOLO('./best.pt')

# 执行预测并获取结果
results = model.predict(source='./valid/images/screenshot-20250504103355_png.rf.71febf531b60b1e416aece6136e28388.jpg')

# 遍历结果
for result in results:
    boxes = result.boxes  # 获取所有检测到的边框
    for box in boxes:
        print("BoundingBox: ", box.xyxy)  # 输出边框坐标(x1, y1, x2, y2)
      
        print("Confidence: ", box.conf)   # 置信度分数
        print("Class: ", box.cls)         # 类别ID
        print("物体是: ",model.names[int(box.cls.item()) ],"范围为: ",box.xyxy.cpu().numpy())