Get Started

Data Preparation

Download Soccer, Tennis, Badminton, Volleyball, and Basketball and put them under <WASB-SBDT_HOME>/datasets.

For Soccer, we provide a setup script. Run cd ./src && sh setup_scripts/setup_soccer.sh .
For Tennis, download Dataset.zip from this link and put it under <WASB-SBDT_HOME> directory. Then unzip the file and rename Dataset directory as tennis .
For Badminton, download TrackNetV2.zip from this link and put it under <WASB-SBDT_HOME> directory. Then use a setup script by running cd ./src && sh setup_scripts/setup_badminton.sh .
For Volleyball, download volleyball_.zip from this link and volleyball_ball_annotations.zip from this link and unzip them. Then put the resulting directories as follows.
For Basketball, download all the zip segments shown here and unzip them to generate NBA_data. Then run a setup script i.e., cd ./src && sh setup_scripts/setup_basketball.sh .

Data structure should be the following:

    datasets
    |-----soccer
    |        └-----videos
    |        └-----frames
    |        └-----annos
    └-----tennis /* renamed from Dataset */
    |        └-----game1
    |        └-----...
    |        └-----game10
    └-----badminton 
    |        └-----match1
    |        └-----...
    |        └-----match26
    |        └-----test_match1
    |        └-----...
    |        └-----test_match3
    └-----volleyball
    |        └-----videos
    |        └-----volleyball_ball_annotation
    └-----basketball /* renamed from NBA_data */
    |        └-----videos
    |        └-----ball-annos
    |
    src

Model Preparation

Download pretrained models listed in MODEL_ZOO.md.

We also provide a setup script to download all the listed models at once. Run cd ./src && sh setup_scripts/setup_weights.sh, then models are located in <WASB-SBDT_HOME>/pretrained_weights.

Evaluation

Here we show the evaluation commands to reproduce the results of Table 2 and Table 3 in our paper.

WASB (Ours, Step=3)

# Soccer
python3 main.py --config-name=eval dataset=soccer model=wasb detector.model_path=../pretrained_weights/wasb_soccer_best.pth.tar

# Tennis
python3 main.py --config-name=eval dataset=tennis model=wasb detector.model_path=../pretrained_weights/wasb_tennis_best.pth.tar

# Badminton
python3 main.py --config-name=eval dataset=badminton model=wasb detector.model_path=../pretrained_weights/wasb_badminton_best.pth.tar

# Volleyball
python3 main.py --config-name=eval dataset=volleyball model=wasb detector.model_path=../pretrained_weights/wasb_volleyball_best.pth.tar

# Basketball
python3 main.py --config-name=eval dataset=basketball model=wasb detector.model_path=../pretrained_weights/wasb_basketball_best.pth.tar

WASB (Ours, Step=1)

# Soccer
python3 main.py --config-name=eval dataset=soccer model=wasb detector.model_path=../pretrained_weights/wasb_soccer_best.pth.tar detector.step=1

# Tennis
python3 main.py --config-name=eval dataset=tennis model=wasb detector.model_path=../pretrained_weights/wasb_tennis_best.pth.tar detector.step=1

# Badminton
python3 main.py --config-name=eval dataset=badminton model=wasb detector.model_path=../pretrained_weights/wasb_badminton_best.pth.tar detector.step=1

# Volleyball
python3 main.py --config-name=eval dataset=volleyball model=wasb detector.model_path=../pretrained_weights/wasb_volleyball_best.pth.tar detector.step=1

# Basketball
python3 main.py --config-name=eval dataset=basketball model=wasb detector.model_path=../pretrained_weights/wasb_basketball_best.pth.tar detector.step=1

MonoTrack [CVPRW2022]

# Soccer
python3 main.py --config-name=eval dataset=soccer model=monotrack detector.postprocessor.use_hm_weight=False detector.model_path=../pretrained_weights/monotrack_soccer_best.pth.tar tracker=intra_frame_peak

# Tennis
python3 main.py --config-name=eval dataset=tennis model=monotrack detector.postprocessor.use_hm_weight=False detector.model_path=../pretrained_weights/monotrack_tennis_best.pth.tar tracker=intra_frame_peak

# Badminton
python3 main.py --config-name=eval dataset=badminton model=monotrack detector.postprocessor.use_hm_weight=False detector.model_path=../pretrained_weights/monotrack_badminton_best.pth.tar tracker=intra_frame_peak

# Volleyball
python3 main.py --config-name=eval dataset=volleyball model=monotrack detector.postprocessor.use_hm_weight=False detector.model_path=../pretrained_weights/monotrack_volleyball_best.pth.tar tracker=intra_frame_peak

# Basketball
python3 main.py --config-name=eval dataset=basketball model=monotrack detector.postprocessor.use_hm_weight=False detector.model_path=../pretrained_weights/monotrack_basketball_best.pth.tar tracker=intra_frame_peak

ResTrackNetV2

# Soccer
python3 main.py --config-name=eval dataset=soccer model=restracknetv2 detector.postprocessor.use_hm_weight=False detector.model_path=../pretrained_weights/restracknetv2_soccer_best.pth.tar tracker=intra_frame_peak

# Tennis
python3 main.py --config-name=eval dataset=tennis model=restracknetv2 detector.postprocessor.use_hm_weight=False detector.model_path=../pretrained_weights/restracknetv2_tennis_best.pth.tar tracker=intra_frame_peak

# Badminton
python3 main.py --config-name=eval dataset=badminton model=restracknetv2 detector.postprocessor.use_hm_weight=False detector.model_path=../pretrained_weights/restracknetv2_badminton_best.pth.tar tracker=intra_frame_peak

# Volleyball
python3 main.py --config-name=eval dataset=volleyball model=restracknetv2 detector.postprocessor.use_hm_weight=False detector.model_path=../pretrained_weights/restracknetv2_volleyball_best.pth.tar tracker=intra_frame_peak

# Basketball
python3 main.py --config-name=eval dataset=basketball model=restracknetv2 detector.postprocessor.use_hm_weight=False detector.model_path=../pretrained_weights/restracknetv2_basketball_best.pth.tar tracker=intra_frame_peak

TrackNetV2 [ICPAI2020]

# Soccer
python3 main.py --config-name=eval dataset=soccer model=tracknetv2 detector.postprocessor.use_hm_weight=False detector.model_path=../pretrained_weights/tracknetv2_soccer_best.pth.tar tracker=intra_frame_peak

# Tennis
python3 main.py --config-name=eval dataset=tennis model=tracknetv2 detector.postprocessor.use_hm_weight=False detector.model_path=../pretrained_weights/tracknetv2_tennis_best.pth.tar tracker=intra_frame_peak

# Badminton
python3 main.py --config-name=eval dataset=badminton model=tracknetv2 detector.postprocessor.use_hm_weight=False detector.model_path=../pretrained_weights/tracknetv2_badminton_best.pth.tar tracker=intra_frame_peak

# Volleyball
python3 main.py --config-name=eval dataset=volleyball model=tracknetv2 detector.postprocessor.use_hm_weight=False detector.model_path=../pretrained_weights/tracknetv2_volleyball_best.pth.tar tracker=intra_frame_peak

# Basketball
python3 main.py --config-name=eval dataset=basketball model=tracknetv2 detector.postprocessor.use_hm_weight=False detector.model_path=../pretrained_weights/tracknetv2_basketball_best.pth.tar tracker=intra_frame_peak

BallSeg [MMSports2019]

# Soccer
python3 main.py --config-name=eval dataset=soccer model=ballseg detector.step=1 detector.postprocessor.use_hm_weight=False detector.model_path=../pretrained_weights/ballseg_soccer_best.pth.tar tracker=intra_frame_peak

# Tennis
python3 main.py --config-name=eval dataset=tennis model=ballseg detector.step=1 detector.postprocessor.use_hm_weight=False detector.model_path=../pretrained_weights/ballseg_tennis_best.pth.tar tracker=intra_frame_peak

# Badminton
python3 main.py --config-name=eval dataset=badminton model=ballseg detector.step=1 detector.postprocessor.use_hm_weight=False detector.model_path=../pretrained_weights/ballseg_badminton_best.pth.tar tracker=intra_frame_peak

# Volleyball
python3 main.py --config-name=eval dataset=volleyball model=ballseg detector.step=1 detector.postprocessor.use_hm_weight=False detector.model_path=../pretrained_weights/ballseg_volleyball_best.pth.tar tracker=intra_frame_peak

# Basketball
python3 main.py --config-name=eval dataset=basketball model=ballseg detector.step=1 detector.postprocessor.use_hm_weight=False detector.model_path=../pretrained_weights/ballseg_basketball_best.pth.tar tracker=intra_frame_peak

DeepBall [VISAPP2019]

# Soccer
python3 main.py --config-name=eval dataset=soccer model=deepball detector=deepball detector.model_path=../pretrained_weights/deepball_soccer_best.pth.tar detector.step=1 tracker=intra_frame_peak

# Tennis
python3 main.py --config-name=eval dataset=tennis model=deepball detector=deepball detector.model_path=../pretrained_weights/deepball_tennis_best.pth.tar detector.step=1 tracker=intra_frame_peak

# Badminton
python3 main.py --config-name=eval dataset=badminton model=deepball detector=deepball detector.model_path=../pretrained_weights/deepball_badminton_best.pth.tar detector.step=1 tracker=intra_frame_peak

# Volleyball
python3 main.py --config-name=eval dataset=volleyball model=deepball detector=deepball detector.model_path=../pretrained_weights/deepball_volleyball_best.pth.tar detector.step=1 tracker=intra_frame_peak

# Basketball
python3 main.py --config-name=eval dataset=basketball model=deepball detector=deepball detector.model_path=../pretrained_weights/deepball_basketball_best.pth.tar detector.step=1 tracker=intra_frame_peak

DeepBall-Large

# Soccer
python3 main.py --config-name=eval dataset=soccer model=deepball_large detector=deepball detector.model_path=../pretrained_weights/deepball-large_soccer_best.pth.tar detector.step=1 tracker=intra_frame_peak

# Tennis
python3 main.py --config-name=eval dataset=tennis model=deepball_large detector=deepball detector.model_path=../pretrained_weights/deepball-large_tennis_best.pth.tar detector.step=1 tracker=intra_frame_peak

# Badminton
python3 main.py --config-name=eval dataset=badminton model=deepball_large detector=deepball detector.model_path=../pretrained_weights/deepball-large_badminton_best.pth.tar detector.step=1 tracker=intra_frame_peak

# Volleyball
python3 main.py --config-name=eval dataset=volleyball model=deepball_large detector=deepball detector.model_path=../pretrained_weights/deepball-large_volleyball_best.pth.tar detector.step=1 tracker=intra_frame_peak

# Basketball
python3 main.py --config-name=eval dataset=basketball model=deepball_large detector=deepball detector.model_path=../pretrained_weights/deepball-large_basketball_best.pth.tar detector.step=1 tracker=intra_frame_peak

Training

TBA

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

GET_STARTED.md

GET_STARTED.md

Get Started

Data Preparation

Model Preparation

Evaluation

WASB (Ours, Step=3)

WASB (Ours, Step=1)

MonoTrack [CVPRW2022]

ResTrackNetV2

TrackNetV2 [ICPAI2020]

BallSeg [MMSports2019]

DeepBall [VISAPP2019]

DeepBall-Large

Training

Files

GET_STARTED.md

Latest commit

History

GET_STARTED.md

File metadata and controls

Get Started

Data Preparation

Model Preparation

Evaluation

Training