Popping Dance Style Classification Through Inception V3 Feature Extraction and LSTM Classification
This program classifies Popping Dance Styles through using Inception V3 for feature extraction and LSTM for classification, as proposed by Harvey (2017) in his blog, Five Video Classification Methods. This metholodogy was applied to my own dataset composed of 1310 Popping Dance style videos under 14 classes.
See this link for the full details of the said blog https://medium.com/@harvitronix/five-video-classification-methods-implemented-in-keras-and-tensorflow-99cad29cc0b5
The whole methodology for method #4 (as adopted in this project) is seen in the same link. Refer to this for his github repository Popping-Dance-Style-Classification
Requirements
Keras>=2.0.2 numpy>=1.12.1 pandas>=0.19.2 tqdm>=4.11.2 matplotlib>=2.0.0 Pillow>=2.1.0 h5py>=2.7.0
see requirements.py
ffmpeg https://www.ffmpeg.org/ VideoPad http://www.nchsoftware.com/videopad/index.html
How to Use
This code requires Keras 2 and TensorFlow 1 or greater installed. Please see the requirements.txt file. To ensure you're up to date, run:
pip install -r requirements.txt
Get the dataset
Download the dataset in the data folder. Download the data from https://drive.google.com/open?id=170nlQKM58cw_h-GKBa6YoxzHJLFNnS_i and copy it to data folder
Working with the data
Still in the data folder, unrar e 14PDSC.rar
Create folders with mkdir train && mkdir test && mkdir sequences && mkdir checkpoints
Now, move videos to appropriate folders and extract frames:
python 1_move_files.py
python 2_extract_files.py
Extracting features
Go to the main folder. Extract feature with extract_features.py
Train
To train run train.py. The model is defined at models.py
To see progress while training, run tensorboard --logdir=data/logs
Validate
To validate, run validate_rnn.py
Results
The model resulted to a 93% accuracy and 26% loss.
References
Methodology Citation
Harvey, M. (2017). Five video classification methods implemented in Keras and TensorFlow [Blog Post]. Retrieved from https://blog.coast.ai/five-video-classification -methods-implemented-in-keras-and-tensorflow-99cad29cc0b5