# Training and analysis
For training our models you need to install `requirements.txt` from the main directory
```bash
pip install -r requirements.txt
```
Data used for training is available [here](https://simtk.org/frs/?group_id=1918) (doi:10.18735/j0rz-0k12)
Use the following scripts for training and analyzing models. Files are listed in the order in which they should be run.
| File | Description |
|:------------- |:-------------|
| process_annotations.ipynb* | Processes OpenPose json files |
| process_frames.ipynb* | Normalize pozes in frames |
| combine_video_csvs.ipynb | Combines time series of poses with labels |
| split_ids.ipynb | Training, validation, testing split |
| compute_SEMLS_residuals.ipynb | Builds simple models for SEMLS to control for demographics |
| process_raw_videos.ipynb* | Additional processing of trajectories (missing data, smoothing, orientation) |
| cnn_predict_doublesided_var.ipynb* | Models for variables that depend on the side (GDI, knee flexion at max extension) |
| cnn_predict_singlesided_var.ipynb* | Models for variables that don't depend on the side (cadence, speed) |
| select_optimal_epoch.ipynb | Choose the best model based on validation error |
| calculate_corr_rmse.ipynb | Get performance metrics for classification and regression tasks |
| calculate_SEMLS_ROC.ipynb | Get performance metrics for the binary prediction task (SEMLS) |
| statistical_analysis.ipynb | Analyze results from models |
For predicting on new video data you need to run files with *.
步态提取-使用卷积神经网络从视频中提取行人步态-附项目源码-优质项目实战.zip （460个子文件） 
Fig3a.csv 38KB Fig2a-d.csv 25KB SuppFig2.csv 13KB Fig3b.csv 12KB Fig2e.csv 6KB Fig2g.csv 5KB Fig2f.csv 5KB SuppFig3.csv 344B Dockerfile 265B .dockerignore 21B weights-21-0.8301.hdf5 460KB weights-73-66.9281.hdf5 460KB demo.ipynb 341KB rf_lm_predict.ipynb 17KB cnn_predict_doublesided_var.ipynb 17KB cnn_predict_singlesided_var.ipynb 16KB rr_lm_predict_gdi.ipynb 16KB statistical_analysis.ipynb 14KB compute_SEMLS_residuals.ipynb 8KB calculate_SEMLS_ROC.ipynb 7KB process_raw_videos.ipynb 7KB Figure3.ipynb 5KB calculate_corr_rmse.ipynb 5KB split_ids.ipynb 5KB Figure4.ipynb 4KB process_frames.ipynb 4KB select_optimal_epoch.ipynb 4KB Figure2.ipynb 3KB Figure5.ipynb 3KB process_annotations.ipynb 3KB combine_video_csvs.ipynb 2KB create_all_videos_dict.ipynb 1KB input_000000000345_keypoints.json 2KB input_000000000346_keypoints.json 2KB input_000000000347_keypoints.json 2KB input_000000000149_keypoints.json 1KB input_000000000181_keypoints.json 870B input_000000000182_keypoints.json 870B input_000000000179_keypoints.json 869B input_000000000171_keypoints.json 868B input_000000000174_keypoints.json 868B input_000000000177_keypoints.json 867B input_000000000175_keypoints.json 867B input_000000000176_keypoints.json 867B input_000000000167_keypoints.json 866B input_000000000168_keypoints.json 864B input_000000000173_keypoints.json 864B input_000000000172_keypoints.json 864B input_000000000178_keypoints.json 864B input_000000000170_keypoints.json 863B input_000000000169_keypoints.json 862B input_000000000384_keypoints.json 857B input_000000000380_keypoints.json 856B input_000000000394_keypoints.json 855B input_000000000260_keypoints.json 854B input_000000000294_keypoints.json 854B input_000000000399_keypoints.json 853B input_000000000201_keypoints.json 853B input_000000000256_keypoints.json 853B input_000000000330_keypoints.json 853B input_000000000341_keypoints.json 853B input_000000000029_keypoints.json 853B input_000000000254_keypoints.json 852B input_000000000395_keypoints.json 852B input_000000000374_keypoints.json 852B input_000000000222_keypoints.json 852B input_000000000253_keypoints.json 852B input_000000000328_keypoints.json 852B input_000000000284_keypoints.json 852B input_000000000280_keypoints.json 852B input_000000000136_keypoints.json 852B input_000000000123_keypoints.json 852B input_000000000127_keypoints.json 852B input_000000000015_keypoints.json 852B input_000000000315_keypoints.json 851B input_000000000344_keypoints.json 851B input_000000000017_keypoints.json 851B input_000000000238_keypoints.json 851B input_000000000392_keypoints.json 851B input_000000000302_keypoints.json 851B input_000000000072_keypoints.json 851B input_000000000030_keypoints.json 851B input_000000000234_keypoints.json 851B input_000000000192_keypoints.json 851B input_000000000121_keypoints.json 851B input_000000000018_keypoints.json 851B input_000000000021_keypoints.json 851B input_000000000360_keypoints.json 851B input_000000000296_keypoints.json 851B input_000000000122_keypoints.json 850B input_000000000279_keypoints.json 850B input_000000000235_keypoints.json 850B input_000000000126_keypoints.json 850B input_000000000319_keypoints.json 850B input_000000000124_keypoints.json 850B input_000000000381_keypoints.json 850B input_000000000218_keypoints.json 850B input_000000000388_keypoints.json 850B input_000000000317_keypoints.json 850B input_000000000312_keypoints.json 850B共 460 条
- 1
- 2
- 3
- 4
- 5
资源评论
__AtYou__
- 粉丝: 3512
- 资源: 2175
下载权益
C知道特权
VIP文章
课程特权
VIP享7折，此内容立减5.97元 开通VIP
最新资源
- 动手学深度学习，沐神版配套代码，所有代码均可在jupyter中运行，内附有极为详尽的代码注释
- qaxbrowser-1.1.32574.52.exe (奇安信浏览器windows安装包)
- C#编写modbus tcp客户端读取modbus tcp服务器数据
- 某房地产瑞六补环境部分代码
- 基于Matlab实现无刷直流电机仿真（模型+说明文档）.rar
- AllSort(直接插入排序，希尔排序，选择排序，堆排序，冒泡排序，快速排序，归并排序)
- 模拟qsort,改造冒泡排序使其能排序任意数据类型，即日常练习
- carsim+simulink联合仿真实现变道 包含路径规划算法+mpc轨迹跟踪算法 可选simulink版本和c++版本算法 可以适用于弯道道路，弯道车道保持，弯道变道 carsim内规划轨迹可视化
- 数组经典习题之顺序排序和二分查找和冒泡排序
- 永磁同步电机神经网络自抗扰控制，附带编程涉及到的公式文档，方便理解，模型顺利运行，效果好，位置电流双闭环采用二阶自抗扰控制，永磁同步电机三闭环控制，神经网络控制，自抗扰中状态扩张观测器与神经网络结合
资源上传下载、课程学习等过程中有任何疑问或建议，欢迎提出宝贵意见哦~我们会及时处理！
点击此处反馈
