Request the users to cite the paper when working with the dataset.

Ravi, Aravind, Harshwin Venugopal, Sruthy Paul, and Hamid R. Tizhoosh. 
"A Dataset and Preliminary Results for Umpire Pose Detection Using SVM Classification of Deep Features." 
In 2018 IEEE Symposium Series on Computational Intelligence (SSCI), pp. 1396-1402. IEEE, 2018.

Paper: https://ieeexplore.ieee.org/abstract/document/8628877/
SNOW Dataset (Six, No Ball, Out, Wide)
Dataset: https://drive.google.com/drive/folders/1ljDIz69mJqDzBlUxABP0c34NgdynzWPX 
         https://drive.google.com/open?id=1fXoT_eW-13Qmek4M2jut6ryHBHTzPhqT

*******************************************************************************************************************
Description
************

The dataset is provided in the 'data' folder:

I. data/
  a. umpire_non_umpire - Contains files named umpire_xxx.jpg and non_umpire_xxx.jpg
  b. umpire_poses  - Contains files named no_ball_xxx.jpg, out_xxx.jpg, sixes_xxx.jpg, 
                     wide_xxx.jpg and no_action_xxx.jpg
  
  The train and test data were split as presented in the reference paper.

********************************************************************************************************************
NOTE: Unzip the folders within the same 'data' directory and run the scripts from the 'code' directory.
********************************************************************************************************************

II. code/
  a. inception - Contains code to:
                (i) generate deep features for model 1 and model 2 based on pretrained 
                    inceptionv3 model.
                (ii) train and save SVM classifiers for model 1 and model 2 based on inceptionv3 features.
                (iii) generate the summary from a cricket video based on respective classifier.
  b. vgg19 - Contains code to:
             (i) generate deep features for model 1 and model 2 based on pretrained 
                 vgg19 model.
             (ii) train and save SVM classifiers for model 1 and model 2 based on vgg19 features.
             (iii) generate the summary from a cricket video based on respective classifier.
********************************************************************************************************************
NOTE: The cricket video to geenrate the highlights is not provided in this repo.
********************************************************************************************************************

III. features/
  a. inception - Contains the deep features extracted using the Inception V3 model.
  b. vgg19  - Contains the deep features extracted using the Inception V3 model.

Each folder contains features for the respective models to perform cricket highlights generation.

Model 1 - Classify between an Umpire image and a Non Umpire image (Those images that are umpire will be passed to
          Model 2 for classification in the video summarization pipeline)
Model 2 - Classify between the five Umpire action classes (Six, No Ball, Out, Wide, No Action)

IV. models/
  a. inception_svm - Contains the SVM model 1 and model 2 trained on inceptionV3 features.
  b. vgg19_svm - Contains the SVM model 1 and model 2 trained on vgg19 features.

V. results_summary/
  Contains the output summarized cricket highlights.