ViWi Dataset
Overview
ViWi Scenarios
Applications
ML Competition
ViWi-BT Challenge

The ViWi Dataset Scenarios


1) The ASU Downtonw Scenario (ASUDT1) - (multi-user)


A top and perspective views of the multiuser outdoor scenario "ASUDT1". The top view (on the left) is showing the basestation locations, the x-y directions, and the fields of view of the three cameras installed at each basestation. The perspective view (i.e., birdview) is shown on the right, and it provides an overview of the objects in the environment.


Description of the Scenario

  • Overview: This ASU Downtown 1 scenario, "ASUDT1", is an outdoor scenario with multiple wireless users. It depicts a busy downtown street with its various elements, e.g., cars, buses, trucks, skyscrapers, buildings, lamp posts,...etc. In this scenario, every vehicle is equipped with a mmWave omini-directional wireless receiver. There are two basestations each of which is equipped with a parameterized antenna array, i.e., the size and shape of the antenna array could be customized using the generation script. There are three differently-oriented cameras installed at each basestation. The fields of view of those cameras are overlapping, see the topview photo of the scenario.

  • Key components:

    • Basestations: There are two basestations set at (x,y,z) = (80,14,4.5) and (x,y,z) = (160,-14,4.5). Each one has a 28 GHz mmWave antenna array and three differently-oriented cameras. Cameras 1,2, and 3 are installed at basestation 1, and cameras 4, 5, and 6 are installed at basestation 2. The side cameras (1, 3, 4, and 6) have a field of view of 75 degrees while the central cameras (2 and 5) have a wider view of 110 degrees. The distance between the two basestations is 80 meters, and they have a shared camera field of view. In particular, camera 3 and 4 view the same segment of the street.

    • Car trajectries: Each vehicle drives along the street in one of four lanes. Lanes 1 and 2 have vehicles moving along the negative x-axis, and the center of each is y = 1.875 m and 5.625 m. Lanes 3 and 4 have vehicles moving along the positive x-axis with centers at y = -1.875 m and -5.625 m. A vehicle stays in one lane and only switches once it reaches the end of the street.

    • Vehicles: There is a total of 60 vehicles in the scenario, 52 cars, 3 trucks, and 5 buses. Not all of them are visible at any instance (scene) as they are continuously moving, entering and leaving the environment.

  • Version support: Version 2

Data filees for the ASUDT_28 Scenario

2) The Colocated-Camera Scenario with Blocked View (Single User)


TopView BirdView
A top and perspective views of the colocated-camera scenario with blockages. The top view (on the left) is showing the 5 trajectories of the car, the stationay blockages (buses), the location of the BS, and the fields of view of the three cameras. The perspective view (on the right) shows an oveerview of the street, building, car, buses, and the BS.


Description of the Scenario

  • Overview: This scenario, "colo_cam_blk", is an outdoor scenario depicting a single car driving through a city street with two stationary buses. It has a single Base Station (BS) equipped with a mmWave antenna and 3 cameras.

  • Key components:

    • Base Station: It has one mmWave antenna and three differently-oriented cameras with the following fields of view 75, 100, and 75 degrees, respectively. The BS is set at 5 meters high and in the middle of the street.

    • Car trajeectories: The car goes through one of five trajectories, each of which is 90 meters long and has 1000 equally spaced points (0.089 m). Hence, they, all together, creeate a 5000-point user grid.

  • Version support: Version 1

For more details on the street, building, and other objects, please refer to ViWi: A Deep Learning Dataset Framework for Vision-Aided Wireless Communications

Data filees for the colo_cam_blk Scenario

3) The Colocated-Camera Scenario with Direct View (Single User)


TopView BirdView
A top and perspective views of the colocated-camera scenario. The top view (on the left) is showing the 5 trajectories of the car, the location of the BS, and the fields of view of the three cameras. The perspective view (on the right) shows an oveerview of the street, building, car, and the BS.


Description of the Scenario

  • Overview: This scenario, "colo_cam", is an outdoor scenario depicting a single car driving through a city street. It has a single Base Station (BS) equipped with a mmWave antenna and 3 cameras.

  • Key components:

    • Base Station: It has one mmWave antenna and three differently-oriented cameras with the following fields of view 75, 100, and 75 degrees, respectively. The BS is set at 5 meters high and in the middle of the street.

    • Car trajeectories: The car goes through one of five trajectories, each of which is 90 meters long and has 1000 equally spaced points (0.089 m). Hence, they, all together, creeate a 5000-point user grid.

  • Version support: Version 1

For more details on the street, building, and other objects, please refer to ViWi: A Deep Learning Dataset Framework for Vision-Aided Wireless Communications

Data filees for the colo_cam Scenario

4) The Distributed-Camera Scenario with Blocked View (Single User)


TopView BirdView
A top and perspective views of the distributed-camera scenario with blockage. The top view (on the left) is showing the 5 trajectories of the car, the location of the three BSs, the stationaty buses, and the fields of view of the three cameras. The perspective view (on the right) shows an oveerview of the street, building, car, buses, and the BS.


Description of the Scenario

  • Overview: This scenario, "dist_cam_blk", is an outdoor scenario depicting a single car driving through a city street with two stationary buss. It has three Base Stations (BSs), each is equipped with a mmWave antenna and an RGB camera.

  • Key components:

    • Base Station: It has one mmWave antenna and one camera with a 100-degree field of view. The BSs are set at 5 meters high and 30 meters apart.

    • Car trajeectories: The car goes through one of five trajectories, each of which is 90 meters long and has 1000 equally spaced points (0.089 m). Hence, they, all together, creeate a 5000-point user grid.

  • Version support: Version 1

For more details on the street, building, and other objects, please refer to ViWi: A Deep Learning Dataset Framework for Vision-Aided Wireless Communications

Data filees for the dist_cam_blk Scenario

5) The Distributed-Camera Scenario with Direct View (Single User)


TopView BirdView
A top and perspective views of the distributed-camera scenario. The top view (on the left) is showing the 5 trajectories of the car, the location of the three BSs, the stationaty buses, and the fields of view of the three cameras. The perspective view (on the right) shows an oveerview of the street, building, car, buses, and the BS.


Description of the Scenario

  • Overview: This scenario, "dist_cam", is an outdoor scenario depicting a single car driving through a city street with two stationary buss. It has three Base Stations (BSs), each is equipped with a mmWave antenna and an RGB camera.

  • Key components:

    • Base Station: It has one mmWave antenna and one camera with a 100-degree field of view. The BSs are set at 5 meters high and 30 meters apart.

    • Car trajeectories: The car goes through one of five trajectories, each of which is 90 meters long and has 1000 equally spaced points (0.089 m). Hence, they, all together, creeate a 5000-point user grid.

  • Version support: Version 1

For more details on the street, building, and other objects, please refer to ViWi: A Deep Learning Dataset Framework for Vision-Aided Wireless Communications

Data filees for the dist_cam Scenario

Citation and License

  • In order to use the ViWi datasets/codes or any (modified) part of them, please cite

    1. The ViWi paper: M. Alrabeiah, A. Hredzak, Z. Liu, and A. Alkhateeb,"ViWi: A Deep Learning Dataset Framework for Vision-Aided Wireless Communications" submitted to IEEE Vehicular Technology Conference, Nov. 2019.
      @InProceedings{Alrabeiah19,
      author = {Alrabeiah, M. and Hredzak, A. and Liu, Z. and Alkhateeb, A.},
      title = {ViWi: A Deep Learning Dataset Framework for Vision-Aided Wireless Communications},
      booktitle = {submitted to IEEE Vehicular Technology Conference},
      year = {2019},
      month = {Nov.},
      }
    2. The Remcom Wireless InSite website: Remcom, Wireless insite, https://www.remcom.com/wireless-insite

  • The ViWi dataset is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License Creative Commons License.

Join the ViWi mailing list!

Please provide your contact information to get the latest updates about the ViWi dataset