Skip to content

MitPandya/Human-Action-Recognition-and-Video-Classification-using-SVM-and-Deep-CNN-

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

24 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

ML_Capstone ( Human Action Recognition and Video Classification using SVM and Deep-CNN )

code instructions

Please install requirement packages from requirements.txt file.
Please download data files from this site http://crcv.ucf.edu/data/UCF101/UCF101.rar

SVM

  1. To create train and test data for SVM run the following commands
    python split_data.py train python split_data.py test This will create train and test folder with relevant video frames.

  2. To train and test the above created data run the following command
    python walk.py
    OR

  3. To test the data from already trained model gunzip svm.pkl.gz file to svm.pkl and run python test_from_model.py

  4. To evaluate the results and create heatmap run
    python eval_results.py

CNN

  1. To create train and test data for CNN run the following commands
    python split_data_cnn.py train python split_data_cnn.py test This will create train and test folder with relevant video frames.

  2. To convert our dataset according to Cifar10 format follow following steps
    git clone https://github.com/gskielian/PNG-2-CIFAR10 run ./resize-script.sh to resize image frames to 32*32 rgb pixels now run, python convert-images-to-cifar-format.py this will create batch file for train and test. Copy and paste these files to /tmp/cifar10-data/ folder

  3. Next train cifar10 model (https://github.com/tensorflow/models/tree/master/tutorials/image/cifar10) on our above created dataset
    python cifar10/cifar10_train.py

  4. To test from trained cifar10 model (https://github.com/tensorflow/models/tree/master/tutorials/image/cifar10) try
    python cifar10_eval.py

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published