---
license: mit
task_categories:
- image-segmentation
- image-classification
- object-detection
pretty_name: Semantic segmenation and pose estimation from RGB-D
dataset_info:
  features:
  - name: rgb
    dtype: image
  - name: depth
    dtype: image
  - name: mask
    dtype: image
  - name: meta
    list:
    - name: colors
      sequence: float64
    - name: file
      dtype: string
    - name: id
      dtype: int64
    - name: model
      dtype: string
    - name: numberOfColors
      dtype: int64
    - name: orientation
      sequence: float64
    - name: position
      sequence: float64
    - name: positionOnImage
      sequence: int64
    - name: size
      sequence: float64
    - name: sizeOnImage
      sequence: int64
  splits:
  - name: train
    num_bytes: 3340733260.96
    num_examples: 1106
  download_size: 3319212411
  dataset_size: 3340733260.96
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
---

RGB-D dataset for instance segmentation (from RGB or depth) and pose estimation of individual objects. Data has been generated by randomizing bin contents in Webots. 
Each instance contains a mask image as well meta data containing labels, position, and size of each object. 

<video src='https://cdn-uploads.huggingface.co/production/uploads/655b1b359d249b4ab388d4a2/l6b76ezxkPi6lG3Fr6_kj.mp4' width=720/>

You can create your own data by opening webots_grasp.wbt in the world directory using [Webots](https://www.cyberbotics.com).