movie_shot_classification_dataset

A dataset with classified film shots.

Dataset classes

The dataset contains the following 10 classes that where extracted from movies

Shot type	instances
Static	985
Ηandheld	273
Panoramic	207
Travelling_in	55
Vertical_static	52
Zoom in	51
Aerial	51
Travelling_out	46
Panoramic_lateral	46
Vertical_moving	37
Total	1803

Dataset files

After extracting a sequence of 5 frames/sec for all shots, three kinds of pre-computed features are available in a npy format for the dataset:

1. sequential_vgg_features directory contains, for each shot, a sequence of features that are extracted using the first linear layer of the VGG16 network
2. sequential_features directory contains, for each shot, a sequence of features that are extracted using the visual analysis from multimodal_movie_analysis library.
3. aggregated_features directory contains, for each shot, 6 statistics (mean, std, median by std ratio, top-10 percentile, mean of the delta features and std of the delta features) for the non-object features and the mean values of the object features that are calculated based on a sequence of features that are extracted using the visual analysis from multimodal_movie_analysis library.

Cite

If you are using this dataset, please use the following citation

@inproceedings{petrogianni2022film,
  title={Film Shot Type Classification Based on Camera Movement Styles},
  author={Petrogianni, Antonia and Koromilas, Panagiotis and Giannakopoulos, Theodoros},
  booktitle={Iberian Conference on Pattern Recognition and Image Analysis},
  pages={602--615},
  year={2022},
  organization={Springer}
}