README.md

---
license_link: LICENSE.md
---

This dataset contains 1024 60 second video clips of Bleeding Edge gameplay (75GB). The data has already been processed into the following format: 300x180 videos sampled at 10 fps.

# Dataset Structure

## Data Files

- testing_dataset_part1.zip & testing_dataset_part2.zip – Contains all 1024 60 second trajectories used for our evaluation.

4 examples from the dataset:
- FB[…].npz – .npz file (described below)
- FB[…].mp4 – 60 seconds .mp4 video of the images from the .npz file. Provided for convenience.

## Data Instances

A typical data entry in the dataset consists of a .npz file that contains uncompressed video frames and corresponding human actions, and accompanying .mp4 file of the same content for easier viewing.

Below is an example from the dataset:

```bash
  >>> example_npz["actions"].shape
  (610, 16)

  >>> example_npz["actions"][:2]
  array([[ 1.,  0.,  0.,  0.,  0.,  0.,  0.,  0.,  0.,  0.,  0.,  0.,  6., 10.,  5.,  5.], 
         [ 1.,  0.,  0.,  0.,  0.,  0.,  0.,  0.,  0.,  0.,  0.,  0.,  6., 10.,  5.,  5.]], dtype=float32)

  >>> example_npz["states"].shape
  (610, 56)

  >>> example_npz["states"][:2]
  array([[ 0.5925019 ,  0.39726806,  0.478     ,  0.        ,  1.        ,
           1.        ,  0.02      ,  0.04      , -1.        , -1.        ,
          -1.        , -1.        , -1.        , -1.        , -1.        ,
          -1.        , -1.        , -1.        , -1.        , -1.        ,
          -1.        , -1.        , -1.        , -1.        , -1.        ,
          -1.        , -1.        , -1.        , -1.        , -1.        ,
          -1.        , -1.        , -1.        , -1.        , -1.        ,
          -1.        , -1.        , -1.        , -1.        , -1.        ,
          -1.        , -1.        , -1.        , -1.        , -1.        ,
          -1.        , -1.        , -1.        , -1.        , -1.        ,
          -1.        , -1.        , -1.        , -1.        , -1.        ,
          -1.        ],
         [ 0.59672105,  0.39742166,  0.478     ,  0.        ,  1.        ,
           1.        ,  0.02      ,  0.04      , -1.        , -1.        ,
          -1.        , -1.        , -1.        , -1.        , -1.        ,
          -1.        , -1.        , -1.        , -1.        , -1.        ,
          -1.        , -1.        , -1.        , -1.        , -1.        ,
          -1.        , -1.        , -1.        , -1.        , -1.        ,
          -1.        , -1.        , -1.        , -1.        , -1.        ,
          -1.        , -1.        , -1.        , -1.        , -1.        ,
          -1.        , -1.        , -1.        , -1.        , -1.        ,
          -1.        , -1.        , -1.        , -1.        , -1.        ,
          -1.        , -1.        , -1.        , -1.        , -1.        ,
          -1.        ]], dtype=float32)

  >>> example_npz["images"].shape
  (610, 3, 180, 300)

  >>> example_npz["images"][0]
  array([[[185, 183, 183, ..., 121, 120, 121],
          [200, 199, 200, ..., 106, 105, 103],
          [200, 200, 202, ...,  96,  96,  91],
          ...,
          [159, 160, 149, ..., 128, 127, 136],
          [157, 146, 174, ..., 127, 126, 126],
          [121, 172, 179, ..., 126, 125, 125]],
  
         [[195, 195, 195, ...,  69,  72,  70],
          [210, 211, 212, ...,  74,  75,  74],
          [210, 213, 214, ...,  77,  77,  77],
          ...,
          [159, 160, 149, ..., 126, 125, 134],
          [157, 146, 174, ..., 127, 125, 125],
          [121, 172, 179, ..., 127, 126, 126]],
  
         [[201, 199, 201, ...,  45,  45,  44],
          [216, 213, 216, ...,  49,  48,  48],
          [216, 214, 218, ...,  52,  50,  50],
          ...,
          [145, 146, 134, ..., 108, 107, 116],
          [142, 132, 160, ..., 108, 107, 107],
          [107, 158, 164, ..., 108, 107, 107]]], dtype=uint8)
```

An example of an image from the dataset: ![Example Image](example_img.png)

# Dataset Creation

We worked with the game studio Ninja Theory and their game [Bleeding Edge](https://www.bleedingedge.com/) – a 3D, 4v4 multiplayer video game. Recorded data was "replayed" in a special build of the game to render and export it in the form of 300x180 images and corresponding controller actions.

# Source Data

The source data was collected by Ninja Theory, a Xbox Game Studio. 

# Data Collection and Processing

The human gameplay data collection and processing was conducted with approval from the Microsoft Institutional Review Board (IRB).

Player data was deidentified prior to use in this research project.

# Limitations

This gameplay data represents gameplay on a single level  of the game Bleeding Edge. 
This dataset represents gameplay recordings from a variety of skilled and unskilled gameplayers, representing diverse demographic characteristics. Not all possible player characteristics are represented and model performance may therefore vary.
Images in the dataset include watermark and provenance metadata. Do not remove the watermark or provenance metadata.   

# License

The model is licensed under the [Microsoft Research License](LICENSE.md).

This work has been funded by Microsoft Research.

# Privacy & Ethics Statement

[Microsoft Privacy Statement](https://go.microsoft.com/fwlink/?LinkId=521839)

# Trademark Notice

**Trademarks** This project may contain trademarks or logos for projects, products, or services. Authorized use of Microsoft trademarks or logos is subject to and must follow [Microsoft’s Trademark & Brand Guidelines](https://www.microsoft.com/en-us/legal/intellectualproperty/trademarks/usage/general). Use of Microsoft trademarks or logos in modified versions of this project must not cause confusion or imply Microsoft sponsorship. Any use of third-party trademarks or logos are subject to those third-party’s policies.

# Contact Information

For questions please email to msr-be-data@microsoft.com