README.md

metadata

license: mit
task_categories:
  - object-detection
  - image-segmentation
  - robotics
dataset_info:
  features:
    - name: scene_id
      dtype: string
    - name: image_id
      dtype: string
    - name: obj_id
      dtype: int64
    - name: pose
      sequence:
        sequence: float64
    - name: camera_intrinsics
      sequence:
        sequence: float64
    - name: depth_scale
      dtype: float64
    - name: bbox
      sequence: int64
    - name: visibility
      dtype: float64
    - name: split
      dtype: string
    - name: rgb
      dtype: image
    - name: depth
      dtype: image
    - name: mask
      dtype: image
    - name: mask_visib
      dtype: image
  splits:
    - name: test
      num_bytes: 12240185177.56
      num_examples: 12247
    - name: train
      num_bytes: 8947085481.56
      num_examples: 10222
  download_size: 7105758283
  dataset_size: 21187270659.12
configs:
  - config_name: default
    data_files:
      - split: test
        path: data/test-*
      - split: train
        path: data/train-*

IndustryShapes

Project Page | Paper

IndustryShapes is a new benchmark dataset tailored for 6D object pose estimation in industrial settings. Targeting the challenges of textureless objects, reflective surfaces, and complex assembly tools, this dataset provides high-quality RGB-D data with precise annotations to advance the state of the art in robotic manipulation.

Dataset Features

Unlike traditional datasets focused on household products, IndustryShapes introduces five new industry-relevant object types with challenging properties. The dataset features:

• Realistic Settings: Objects captured in authentic industrial assembly environments.
• Diverse Complexity: Scenes ranging from simple to challenging, including single and multiple objects, as well as multiple instances of the same object.
• Unique Modalities: It is the first dataset to offer RGB-D static onboarding sequences to support model-free and sequence-based approaches.
• Comprehensive Annotations: Includes high-quality annotated poses, bounding boxes, and segmentation masks.

Dataset Organization

The dataset is organized into two parts:

• Classic Set: The Classic Set supports instance-level pose estimation with 21 scenes (13 train, 8 test). Includes images from real industrial scenes with varying complexity, Lab captured and Synthetically generated data.
• Extended Set: Inlucdes three challenging office scenes with unconstrained lighting, distractors, occlusions and diverse viewpoints featuring all objects. It also includes 10 RGB-D static onboarding sequences (2 per object).

Tasks

• 6D Object Pose Estimation (Instance-level and Novel Object)
• Object Detection
• Image Segmentation
• Robotic Manipulation