POQUITO Dataset that accompanies the paper:
"Tiny Satellites, Big Challenges: A Feasibility Study of Machine Vision Pose Estimation for PocketQubes during Conjunctions"
Niki Sajjad, Andrew Price, Mehran Mirshams and Mathieu Salzmann
September 2024

The dataset formatting is adapted from the BOP format. 

The dataset is organized as:
POQUITO_Dataset
|---1_Distance_Check_Data
    |---Training
    |---Validation
    |---Testing
|---2_Performance_Check_Data
    |---Training
    |---Validation
    |---Testing
        |---1_Stationary_cases
        |---2_Sequential_Trajectories
            |---Trajectory1
            |---Trajectory2
            |---Trajectory3
            |---Trajectory4
            |---Trajectory5           
|---3_Satellite_Model
    |---POQUITO_bbox.json
    |---POQUITO.ply
|---4_Earth_Background
    |---real_Earth_images
    |---real_Earth_images_cropped
    |---rendered_Earth_images
|---README.txt

[1] Contains images of the POQUITO target centred on the image. The distance from the camera to the POQUITO is slowly increased.

[2] Contains rgb images with various post processing affects applied.
    -foc- camera depth of field (focal) blur applied as Gaussian blurring.
    -mot- motion blur applied with custom XiaolinWu filter kernels.
    -earth- images of the Earth inserted into the background.
  Testing contains two separate cases, <1_Stationary_cases> and <2_Sequential_Trajectories>
    /Testing/1_Stationary_cases/ is similar to Training and <Validation
    /Testing/2_Sequential_Trajectories/ contains 5 trajectories, the motion blur and focal blur included in these trajectories is physically proportional and is labeled.
      Ex. /2_Performance_Check_Data/Testing/2_Sequential_Trajectories/Trajectory1/rgb_foc/0.45 contains images with a Gaussian blur std of 0.45 applied.
      Ex. /2_Performance_Check_Data/Testing/2_Sequential_Trajectories/Trajectory1/rgb_motEarth/15 contains images with Earth background and a motion blur kernel of 15 applied.

[3] Contains the POQUITO CAD file and POQUITO_bbox.json which contains 8 xyz points [millimetres] defining the bounding box of POQUITO.

[4] Contains the Earth background images inserted during post processing to 2_Performance_Check_Data

rgb folders contain the images
mask_visib folders contain masks indicating the visible portion of the target
scene_gt.json contains the ground truth pose annotations for images in the associated directory.
scene_camera.json contains the camera intrinsics for images in the associated directory. Note we use the same camera intrinsics for the entire dataset. 



If you use this dataset, please consider citing our work.
@article{POQUITO_Sajjad2024,
    title={Tiny Satellites, Big Challenges:A Feasibility Study of Machine Vision Pose Estimation for PocketQubes during Conjunctions},
    author={Niki Sajjad and Andrew Price and Mehran Mirshams and Mathieu Salzmann},
    journal={},
    year={},
    volume={},
    number={},
    pages={},
    doi={}
}