Open chg0901 opened 1 year ago
We recommend using English or English & Chinese for issues so that we could have broader discussion.
https://github.com/JeffDing/mmlabcamp/tree/main/第二期
Homework1.ipynb 课程的所有记录
图像及输出目录在mmpose/output目录下
pth文件存在百度网盘: 链接: https://pan.baidu.com/s/1p4xij8m3byJAl-ZzAXPHrQ?pwd=rzim 提取码: rzim
https://github.com/xixihic/openmmlabhomework/tree/master 本次训练的配置文件在mmpose/config目录下 输出结果在mmpose/result目录下 测试指标在mmpose/test目录下 由于无法拍摄可用的耳朵照片及视频,所以使用网上的照片进行测试,无视频
详见测试数据和结果文件夹 MyEar目录和README.md
best epoch: 196/200
coco/bbox_mAP: 0.8080 coco/bbox_mAP_50: 0.9700 coco/bbox_mAP_75: 0.9700
coco/bbox_mAP_s: -1.0000 coco/bbox_mAP_m: -1.0000 coco/bbox_mAP_l: 0.8080
best epoch: 255/300
coco/AP: 0.740501 coco/AP .5: 1.000000 coco/AP .75: 0.968647 coco/AP (M): -1.000000 coco/AP (L): 0.740501
coco/AR: 0.780952 coco/AR .5: 1.000000 coco/AR .75: 0.976190 coco/AR (M): -1.000000 coco/AR (L): 0.780952
PCK: 0.975057 AUC: 0.137925 NME: 0.040603
单张图片和两个测试视频
RTMDet-tiny模型权重 best_coco_bbox_mAP_epoch_196.pth
RTMPose-s模型权重 best_PCK_epoch_255.pth
对应的模型轻量化转换权重
【七班】MMPose代码实践与耳朵穴位数据集实战【OpenMMLab AI实战营第二期Day3】 【CSDN】Version
推理结果:作业/mmpose/outputs mmdet训练结果:mmdetection mmpose训练结果:mmpose
DONE (t=0.02s).
Average Precision (AP) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.537
Average Precision (AP) @[ IoU=0.50 | area= all | maxDets=100 ] = 0.954
Average Precision (AP) @[ IoU=0.75 | area= all | maxDets=100 ] = 0.501
Average Precision (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = -1.000
Average Precision (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = -1.000
Average Precision (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.537
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 1 ] = 0.598
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 10 ] = 0.650
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.650
Average Recall (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = -1.000
Average Recall (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = -1.000
Average Recall (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.650
06/04 21:46:45 - mmengine - INFO - bbox_mAP_copypaste: 0.537 0.954 0.501 -1.000 -1.000 0.537
06/04 21:46:45 - mmengine - INFO - Epoch(val) [50][2/2] coco/bbox_mAP: 0.5370 coco/bbox_mAP_50: 0.9540 coco/bbox_mAP_75: 0.5010 coco/bbox_mAP_s: -1.0000 coco/bbox_mAP_m: -1.0000 coco/bbox_mAP_l: 0.5370 data_time: 2.0657 time: 2.1472
DONE (t=0.00s).
Average Precision (AP) @[ IoU=0.50:0.95 | area= all | maxDets= 20 ] = 0.741
Average Precision (AP) @[ IoU=0.50 | area= all | maxDets= 20 ] = 1.000
Average Precision (AP) @[ IoU=0.75 | area= all | maxDets= 20 ] = 0.970
Average Precision (AP) @[ IoU=0.50:0.95 | area=medium | maxDets= 20 ] = -1.000
Average Precision (AP) @[ IoU=0.50:0.95 | area= large | maxDets= 20 ] = 0.741
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 20 ] = 0.793
Average Recall (AR) @[ IoU=0.50 | area= all | maxDets= 20 ] = 1.000
Average Recall (AR) @[ IoU=0.75 | area= all | maxDets= 20 ] = 0.976
Average Recall (AR) @[ IoU=0.50:0.95 | area=medium | maxDets= 20 ] = -1.000
Average Recall (AR) @[ IoU=0.50:0.95 | area= large | maxDets= 20 ] = 0.793
06/05 01:37:52 - mmengine - INFO - Evaluating PCKAccuracy (normalized by ``"bbox_size"``)...
06/05 01:37:52 - mmengine - INFO - Evaluating AUC...
06/05 01:37:52 - mmengine - INFO - Evaluating NME...
06/05 01:37:52 - mmengine - INFO - Epoch(val) [200][6/6] coco/AP: 0.740731 coco/AP .5: 1.000000 coco/AP .75: 0.970297 coco/AP (M): -1.000000 coco/AP (L): 0.740731 coco/AR: 0.792857 coco/AR .5: 1.000000 coco/AR .75: 0.976190 coco/AR (M): -1.000000 coco/AR (L): 0.792857 PCK: 0.975057 AUC: 0.141893 NME: 0.039382 data_time: 0.408166 time: 0.432510
06/05 01:37:52 - mmengine - INFO - The previous best checkpoint /data/run01/scz0brk/openmmlab/mmpose/work_dirs/rtmpose-s-ear/best_PCK_epoch_170.pth is removed
先放个作业地址https://github.com/shaohua-pan/openmmlab-hw
在出差,先放个作业地址https://github.com/hjy-pan/openmmlab-hjy
大佬您好,很精彩的实战。我也在尝试耳穴识别,但是您发的耳朵穴位关键点检测数据集链接已过期,能否在文章更新下或私信发下新的下载链接,感谢!
题目:基于RTMPose的耳朵穴位关键点检测
背景:
根据中医的“倒置胎儿”学说,耳朵的穴位反映了人体全身脏器的健康,耳穴按摩可以缓解失眠多梦、内分泌失调等疾病。耳朵面积较小,但穴位密集,涉及耳舟、耳轮、三角窝、耳甲艇、对耳轮等三维轮廓,普通人难以精准定位耳朵穴位。
任务
【4-7任务完成提交】 提交测试集评估指标(不能低于baseline指标的50%) 并使用训练的模型对自己的数据进行预测并保存呈现结果(可采用notebook或者输出图片和视频的形式呈现)。
【参考提交方法和格式】 【提交链接】issue 24 comment 【作业目录】2.Basic_mmdet3.x_V2
PS:任务1,2,3,示例代码作者子豪兄已经提供
目标检测Baseline模型(RTMDet-tiny)![image](https://github.com/open-mmlab/OpenMMLabCamp/assets/8240984/378e8592-abed-4f6b-a955-531a0db6a3f6)
关键点检测Baseline模型(RTMPose-s)![image](https://github.com/open-mmlab/OpenMMLabCamp/assets/8240984/2af4dd12-db79-4b3f-b8a4-b75607855825)
数据集
耳朵穴位关键点检测数据集,MS COCO格式,划分好了训练集和测试集,并写好了样例config配置文件 链接: https://pan.baidu.com/s/1swTLpArj7XEDXW4d0lo7Mg 提取码: 741p 标注人:张子豪、田文博![image](https://github.com/open-mmlab/OpenMMLabCamp/assets/8240984/82bc4ee2-ca64-4959-9461-9a12ba244b5e)
提交方式
请将作业内容上传到你自己的github仓库,并把对应的链接回复在评论区