网站 首页 关键词,个人无网站怎样做cps广告,推荐做微商海报的网站,c 做网站教程这里写目录标题 1.目标检测 Detection2.实例分割 segment3.图像分类 classify4.关键点估计 Keypoint detection5.视频帧检测 video detect6.视频帧分类 video classify7.旋转目标检测 obb detect8.替换yolo11模型 给我点个赞吧#xff0c;谢谢了附录coco80类名称 笔记本 华为m… 这里写目录标题 1.目标检测 Detection2.实例分割 segment3.图像分类 classify4.关键点估计 Keypoint detection5.视频帧检测 video detect6.视频帧分类 video classify7.旋转目标检测 obb detect8.替换yolo11模型 给我点个赞吧谢谢了附录coco80类名称 笔记本 华为matebook14swindows系统cpu 1.装Label-studio 2.装Label-studio-ml-backend 3.装ultralytics 4.装docker desktop 并点击启动
配置好docker-composel.yml文件32 33行 32表示从docker容器里访问容器外的网址label-studio默认端口8080 33表示label-studio API KEY 获取方式 - LABEL_STUDIO_URLhttp://host.docker.internal:8080- LABEL_STUDIO_API_KEYd3ece86209a6a0ca850d468d6c42fa3d7d78be47点击label-studio头像-》点击Account settings-》复制access token
然后拉取镜像第一次耗时一个小时左右。记得科学上网呦
cd label_studio_ml\examples\yolo\
docker-compose up --build 结果如下就表示启动docker成功 在label-studio 后台model处导入label-studio-ml-backend默认网址 http://localhost:9090 如果连接成功会有测试通过显示connected否则报错 自行修改参考docker-compose.yml第46行 ports:- 9090:90901.目标检测 Detection
导入示例标注配置
ViewImage nameimage value$image/RectangleLabels namelabel toNameimage model_score_threshold0.25 opacity0.1Label valuePerson backgroundred/Label valueCar backgroundblue//RectangleLabels
/View最后成功 2.实例分割 segment
替换分割模型 只需要在标签处修改
ViewImage nameimage value$image/PolygonLabels namelabel toNameimage model_score_threshold0.25 opacity0.1Label valueCar backgroundblue/Label valuePerson backgroundred//PolygonLabels
/View然后删除目标检测的预测框就可以 选中图片-》点击左上角 6 Tasks-》Delete Predictions
然后点击随便一张图片重新预测结果
3.图像分类 classify
替换图像分类的标签
ViewImage nameimage value$image/Choices namechoice toNameimage model_score_threshold0.25Choice valueAirplane predicted_valuesaircraft_carrier,airliner,airship,warplane/Choice valueCar predicted_valueslimousine,minivan,jeep,sports_car,passenger_car,police_van//Choices
/View结果显示在左下角的分类里。
4.关键点估计 Keypoint detection
替换标签
ViewRectangleLabels namekeypoints_bbox toNameimage model_skiptrueLabel valueperson//RectangleLabelsKeyPointLabels namekeypoints toNameimagemodel_score_threshold0.75 model_point_threshold0.5 model_add_bboxestrue model_point_size1model_pathyolov8n-pose.ptLabel valuenose predicted_valuesperson model_index0 backgroundred /Label valueleft_eye predicted_valuesperson model_index1 backgroundyellow /Label valueright_eye predicted_valuesperson model_index2 backgroundyellow /Label valueleft_ear predicted_valuesperson model_index3 backgroundpurple /Label valueright_ear predicted_valuesperson model_index4 backgroundpurple /ViewLabel valueleft_shoulder predicted_valuesperson model_index5 backgroundgreen /Label valueleft_elbow predicted_valuesperson model_index7 backgroundgreen /Label valueleft_wrist predicted_valuesperson model_index9 backgroundgreen /Label valueright_shoulder predicted_valuesperson model_index6 backgroundblue /Label valueright_elbow predicted_valuesperson model_index8 backgroundblue /Label valueright_wrist predicted_valuesperson model_index10 backgroundblue //ViewViewLabel valueleft_hip predicted_valuesperson model_index11 backgroundbrown /Label valueleft_knee predicted_valuesperson model_index13 backgroundbrown /Label valueleft_ankle predicted_valuesperson model_index15 backgroundbrown /Label valueright_hip predicted_valuesperson model_index12 backgroundorange /Label valueright_knee predicted_valuesperson model_index14 backgroundorange /Label valueright_ankle predicted_valuesperson model_index16 backgroundorange //View/KeyPointLabelsImage nameimage value$image /
/View展示结果
5.视频帧检测 video detect
标签
ViewVideo namevideo value$video/VideoRectangle namebox toNamevideo model_trackerbotsort model_conf0.25 model_iou0.7 /Labels namelabel toNamevideoLabel valuePerson backgroundred/Label valueCar backgroundblue//Labels
/View第一次处理视频会比较长因为他是完整的预测完才加载后台可以显示当前处理到多少frame 展示效果如下
6.视频帧分类 video classify
标签
ViewVideo namevideo value$video/TimelineLabels namelabel toNamevideo model_trainablefalse model_score_threshold0.25Label valueBall predicted_valuessoccer_ball /Label valuehamster //TimelineLabels
/View测试失败
7.旋转目标检测 obb detect
测试失败
8.替换yolo11模型
下载好然后放到models目录下 修改 \label-studio-ml-backend\label-studio-ml-backend-master\label_studio_ml\examples\yolo\requirements.txt 把ultralytics更新为 ultralytics~8.3.20 否则不支持yolo11
重启docker
docker-compose down
docker-compose up --build就可以了 记得替换标签时加入model_path“yolo11n.pt” 例如目标检测
ViewImage nameimage value$image/PolygonLabels namelabel toNameimage model_score_threshold0.25 opacity0.1 model_pathyolo11n.ptLabel valueCar backgroundblue/Label valuePerson backgroundred//PolygonLabels
/View实测下来 yolo11n.pt yolo11n-seg.pt yolo11n-pose.pt yolo11n-cls.pt 都能用
给我点个赞吧谢谢了
附录coco80类名称
为了方便大家修改标签信息我附上coco数据集80类名称自行参考
person人
bicycle自行车
car轿车
motorcycle摩托车
airplane飞机
bus公共汽车
train火车
truck卡车
boat船
traffic light交通灯
fire hydrant消防栓
stop sign停车标志
parking meter停车收费表
bench长凳
bird鸟
cat猫
dog狗
horse马
sheep羊
cow牛
elephant大象
bear熊
zebra斑马
giraffe长颈鹿
backpack背包
umbrella雨伞
handbag手提包
tie领带
suitcase手提箱
frisbee飞盘
skis滑雪板
snowboard滑雪单板
sports ball体育用球
kite风筝
baseball bat棒球棒
baseball glove棒球手套
skateboard滑板
surfboard冲浪板
tennis racket网球拍
bottle瓶子
wine glass酒杯
cup杯子
fork叉子
knife刀
spoon勺子
bowl碗
banana香蕉
apple苹果
sandwich三明治
orange橙子
broccoli西兰花
carrot胡萝卜
hot dog热狗
pizza披萨
donut甜甜圈
cake蛋糕
chair椅子
couch长沙发
potted plant盆栽
bed床
dining table餐桌
toilet马桶
tv电视
laptop笔记本电脑
mouse鼠标
remote遥控器
keyboard键盘
cell phone手机
microwave微波炉
oven烤箱
toaster烤面包机
sink水槽
refrigerator冰箱
book书
clock时钟
vase花瓶
scissors剪刀
teddy bear泰迪熊
hair drier吹风机
toothbrush牙刷
