Skip to content

dohbless/Continuous_Sign_Language_Sentence_Based_on_Kinect

Repository files navigation

Continuous-Sign-Language-Sentence-Based-on-Kinect

关键动作的识别

深度学习入门

制作自己的数据集

- TFRecords读取数据方式

https://www.cnblogs.com/xianhan/p/9146485.html
https://blog.csdn.net/rookie_wei/article/details/81275663

简易数据集

https://blog.csdn.net/jesmine_gu/article/details/81155787

神经网络训练

https://blog.csdn.net/dy_guox/article/details/79111949

  • python xmltocsv.py
  • Create train data: python generatetfrecords.py --csv_input=data/dataset_train.csv --output_path=train.record python xmltorecod.py --csv_input=data/dataset_train.csv --output_path=train.record python xmltorecod.py --csv_input=data/test_2.csv --output_path=test.record python xmltorecod.py --csv_input=data/train_4.csv --output_path=train.record -(!!!一定要将main里面的路径改为自己的图片们)对应文件夹的images下
  • change config ssd_mobilenet_v1_coco.config
  • start train

From the tensorflow/models/research/ directory

python ./legacy/train.py --logtostderr --train_dir=training2/ --pipeline_config_path=training2/ssd_mobilenet_v1_coco.config --num_train_steps=20000 --num_eval_steps=2000

生成.pb文件

python export_inference_graph.py \ --input_type image_tensor \ --pipeline_config_path training2/ssd_mobilenet_v1_coco.config \ --trained_checkpoint_prefix training2/model.ckpt-30000 \ --output_directory hand_graph

测试模型并输出

About

No description, website, or topics provided.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published