AC-SUM-GAN: Connecting Actor-Critic and Generative Adversarial Networks for Unsupervised Video Summarization
- From "AC-SUM-GAN: Connecting Actor-Critic and Generative Adversarial Networks for Unsupervised Video Summarization" (IEEE Transactions on Circuits and Systems for Video Technology (IEEE TCSVT 2020), Early Access)
- Written by Evlampios Apostolidis, Eleni Adamantidou, Alexandros I. Metsai, Vasileios Mezaris and Ioannis Patras
- This software can be used for training a deep learning architecture that embeds an Actor-Critic model into a Generative Adversarial Network for automatic video summarization. Training is performed in a fully unsupervised manner without the need for ground-truth data (such as human-generated video summaries). After being unsupervisingly trained on a collection of videos, the AC-SUM-GAN model is capable of producing representative summaries for unseen videos, according to a user-specified time-budget about the summary duration.
- Python 3.6
- PyTorch 1.0.1
Structured h5 files with the video features and annotations of the SumMe and TVSum datasets are available within the "data" folder. The GoogleNet features of the video frames were extracted by Ke Zhang and Wei-Lun Chao and the h5 files were obtained from Kaiyang Zhou. These files have the following structure:
/key /features 2D-array with shape (n_steps, feature-dimension) /gtscore 1D-array with shape (n_steps), stores ground truth improtance score (used for training, e.g. regression loss) /user_summary 2D-array with shape (num_users, n_frames), each row is a binary vector (used for test) /change_points 2D-array with shape (num_segments, 2), each row stores indices of a segment /n_frame_per_seg 1D-array with shape (num_segments), indicates number of frames in each segment /n_frames number of frames in original video /picks positions of subsampled frames in original video /n_steps number of subsampled frames /gtsummary 1D-array with shape (n_steps), ground truth summary provided by user (used for training, e.g. maximum likelihood) /video_name (optional) original video name, only available for SumMe dataset
Original videos and annotations for each dataset are also available in the authors' project webpages:
- TVSum dataset: https://github.com/yalesong/tvsum
- SumMe dataset: https://gyglim.github.io/me/vsum/index.html#benchmark
To train the model using one of the aforementioned datasets and for a number of randomly created splits of the dataset (where in each split 80% of the data is used for training and 20% for testing) use the corresponding JSON file that is included in the "data/splits" directory. This file contains the 5 randomly generated splits that were utilized in our experiments.
For training the model using a single split, run:
python main.py --split_index N (with N being the index of the split)
Alternatively, to train the model for all 5 splits, use the 'run_splits.sh' script and do the following:
chmod +x run_splits.sh # Makes the script executable. ./run_splits # Runs the script.
Please note that after each training epoch the algorithm performs an evaluation step, using the trained model to compute the importance scores for the frames of each video of the test set. These scores are then used by the provided evaluation scripts to assess the overal performance of the model (in F-Score).
The progress of the training can be monitored via the TensorBoard platform and by:
- opening a command line (cmd) and running: tensorboard --logdir=/path/to/log-directory --host=localhost
- opening a browser and pasting the returned URL from cmd
Setup for the training process:
- In 'data_loader.py', specify the path to the 'h5' file of the dataset and the path to the 'json' file containing data about the created splits.
- In 'configs.py', define the directory where the models will be saved to.
Arguments in 'configs.py':
--video_type: The used dataset for training the model. Can be either 'TVSum' or 'SumMe'. --input_size: The size of the input feature vectors (1024 for GoogLeNet features). --hidden_size: The hidden size of the LSTM units. --num_layers: The number of layers of each LSTM network. --regularization_factor: The value of the regularization factor (ranges from 0.0 to 1.0 with a step equal to 0.1). --entropy_coef: The entropy regularization coefficient delta (0.1 in this implementation). --n_epochs: Number of training epochs. --clip: The gradient clipping parameter. --lr: Learning rate. --discriminator_lr: Discriminator's learning rate. --split_index: The index of the current split. --action_state_size: The size of the action-state space (60 in this implementation).
For the parameters with no explicitly defined default values, please read the paper ("Implementation Details" section) or check the 'configs.py' file.
The utilized model selection criterion, that relies on the optimization of core factors of the training process (i.e., the received reward from the Discriminator and the loss that is associated to the training of the Actor), enables the selection of a well-trained model by indicating both the training epoch and the value of the regularization factor. A basic prerequisite for using this criterion is to train the AC-SUM-GAN architecture for all the different values of the regularization factor. To evaluate the trained models of the architecture using different values of the regularization factor and automatically select a well-trained model, run 'pipeline.sh'. To run this file, specify "path_to_experiment" by defining the path to the main folder where the analysis results are stored. For further details about the adopted structure of directories in our implementation, please check line #15 of 'evaluate_exp.sh'.
If you find this code useful in your work, please cite the following publication:
E. Apostolidis, E. Adamantidou, A. I. Metsai, V. Mezaris and I. Patras, "AC-SUM-GAN: Connecting Actor-Critic and Generative Adversarial Networks for Unsupervised Video Summarization," in IEEE Transactions on Circuits and Systems for Video Technology.
DOI: https://doi.org/10.1109/TCSVT.2020.3037883
Copyright (c) 2020, Evlampios Apostolidis, Eleni Adamantidou, Alexandros I. Metsai, Vasileios Mezaris, Ioannis Patras / CERTH-ITI. All rights reserved. This code is provided for academic, non-commercial use only. Redistribution and use in source and binary forms, with or without modification, are permitted for academic non-commercial use provided that the following conditions are met:
- Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer.
- Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation provided with the distribution.
This software is provided by the authors "as is" and any express or implied warranties, including, but not limited to, the implied warranties of merchantability and fitness for a particular purpose are disclaimed. In no event shall the authors be liable for any direct, indirect, incidental, special, exemplary, or consequential damages (including, but not limited to, procurement of substitute goods or services; loss of use, data, or profits; or business interruption) however caused and on any theory of liability, whether in contract, strict liability, or tort (including negligence or otherwise) arising in any way out of the use of this software, even if advised of the possibility of such damage.
This work was supported by the European Union Horizon 2020 research and innovation programme under contract H2020-780656 ReTV. The work of Ioannis Patras has been supported by EPSRC under grant No. EP/R026424/1.