← Browse Dataset

WithVision Bin Picking Dataset

early preview
Dataset thumbnail

Dataset Overview

The WithVision Bin Picking Dataset consists of 50,000 synthetic scenes of YCB objects arranged in realistic bin picking scenarios, for use in computer vision and robotic manipulation training.

Images

Each scene contains the following 640x480 images:

Filename Description
rgb.png main image
n.png surface normal
seg.png segmentation mask
bbox.png bounding box visualization
depth.png depth map
pose.png 3D pose visualization

Metadata

Each scene also includes a metadata.json file containing:

Field Description
scene_name unique identifier for the scene
frame_id frame number within the scene
camera detailed camera parameters
objects array of detected objects with annotations
class_name YCB object class identifier
segmentation_color RGB color values for object segmentation
bounding_box 2D bounding box coordinates [x1, y1, x2, y2]
pose_world_to_object 4x4 transformation matrix for 3D pose

Download Dataset

NOTE: This dataset is in active development and is provided as an early preview for researcher feedback. The final dataset will be released in August 2025.

License

This dataset is released under the MIT License. You are free to use, modify, and distribute this dataset for both commercial and non-commercial purposes, subject to the following conditions:

For the full license text, please refer to the LICENSE file included with the dataset download.

Citation

If you use this dataset in your research, please cite:

@misc{ycb_bin_picking_2025,
  title = {WithVision Bin Picking Dataset},
  author = {Paton, Jeremy},
  year = {2025},
  publisher = {WithVision Inc.},
  version = {1.0},
  howpublished = {\url{https://withvision.ai/datasets/ycb-bin-picking}},
  note = {Synthetic dataset for robotic bin picking based on YCB objects}
}

Contact

Questions? Please email jeremy@withvision.ai.