README.md

The instruction for preparing the TTNet dataset

Prepare the dataset

Download the dataset and unzip annotations

cd prepare_dataset
python download_dataset.py
python unzip.py

Extract images from videos

If you want to extract all images from videos, let's execute:

python extract_all_images.py

or alternatively if you have the supported OpenCV version (3.2, 3.4 or 4.x) and Clang (part of LLVM, needed for automatic binding generation) installed in your system, you can use the executable:

./ttnet_extract_images_multithread

As the authors mentioned in the original paper, images are extracted based on frames that has events. To extract frames based on the key annotated events, let's execute:

python extract_selected_images.py

I highly recommend to run the below command to extract frames that will use for the purpose of smooth labeling events (mentioned in the 5.1 part of the paper).

python extract_smooth_labellings.py

Source code structure

${ROOT}
├── dataset/
    ├── test/
        ├── annotations/
        ├── images/
        ├── videos/
    ├── training
        ├── annotations/
            ├── game_1/
                ├── segmentation_masks/
                ├── ball_markup.json
                ├── events_markup.json
        ├── images/
            ├── game_1/
                ├── img_000001.jpg
                ├── img_000002.jpg
                ...
        ├── videos/
            ├── game_1.mp4
            ├── game_2.mp4
            ...
├── prepare_dataset/
        ├── download_dataset.py
        ├── extract_all_images.py
        ├── extract_selected_images.py
        ├── extract_smooth_labellings.py
        ├── unzip.py
        ├── README.md
├── src/
    ├── config
    │   ├── config.py
    │   ├── __init__.py
    ├── data_process
    │   ├── __init__.py
    │   ├── transformation.py
    │   ├── ttnet_dataloader.py
    │   ├── ttnet_dataset.py
    │   ├── ttnet_data_utils.py
    │   └── ttnet_video_loader.py
    ├── losses
    │   ├── __init__.py
    │   ├── losses.py
    ├── models
    │   ├── __init__.py
    │   ├── model_utils.py
    │   ├── multi_task_learning_model.py
    │   ├── TTNet.py
    │   └── unbalanced_loss_model.py
    └── utils
    │   ├── init_paths.py
    │   ├── __init__.py
    │   ├── logger.py
    │   ├── metrics.py
    │   ├── misc.py
    │   ├── post_processing.py
    │   └── train_utils.py
    ├── demo.py
    ├── demo.sh
    ├── main.py
    ├── test_3rd_phase.sh
    ├── test.py
    ├── train_1st_phase.sh
    ├── train_2nd_phase.sh
    ├── train_3rd_phase.sh
    ├── train.sh
├── README.md