Skip to content
Showcase/RT-2 - Robotic Action Annotation
advancedsurvey

RT-2 - Robotic Action Annotation

Robotic manipulation task evaluation and action segmentation based on RT-2 (Brohan et al., CoRL 2023). Annotators evaluate task success, describe actions, rate execution quality, and segment video into action phases.

Q1: Rate your experience12345Q2: Primary use case?ResearchIndustryEducationQ3: Additional feedback

Configuration Fileconfig.yaml

This Potato config reproduces the annotation task. Save it as config.yaml and run potato start config.yaml to try it.

yaml
# RT-2 - Robotic Action Annotation
# Based on Brohan et al., CoRL 2023
# Paper: https://arxiv.org/abs/2307.15818
# Dataset: https://robotics-transformer2.github.io/
#
# This task evaluates robotic manipulation episodes from the RT-2 benchmark.
# Annotators watch a video of a robot performing a task, evaluate the success
# of the execution, describe the actions taken, rate overall quality, and
# segment the video into distinct action phases.
#
# Task Success:
# - Success: The robot completed the task as instructed
# - Partial Success: The robot made progress but did not fully complete the task
# - Failure: The robot failed to make meaningful progress on the task
#
# Action Phases (Video Annotation):
# - Reaching: Robot arm moving toward the target object
# - Grasping: Robot closing gripper on the object
# - Placing: Robot positioning the object at the target location
# - Moving: Robot transporting the object through space
# - Idle: Robot stationary or resetting
#
# Annotation Guidelines:
# 1. Read the task instruction
# 2. Watch the video carefully
# 3. Evaluate task success
# 4. Describe the actions taken by the robot
# 5. Rate the execution quality
# 6. Segment the video into action phases

annotation_task_name: "RT-2 - Robotic Action Annotation"
task_dir: "."

data_files:
  - sample-data.json

item_properties:
  id_key: "id"
  text_key: "text"

output_annotation_dir: "annotation_output/"
output_annotation_format: "json"

port: 8000
server_name: localhost

annotation_schemes:
  - annotation_type: radio
    name: task_success
    description: "Did the robot successfully complete the instructed task?"
    labels:
      - "Success"
      - "Partial Success"
      - "Failure"
    keyboard_shortcuts:
      "Success": "1"
      "Partial Success": "2"
      "Failure": "3"
    tooltips:
      "Success": "The robot fully completed the task as described in the instruction"
      "Partial Success": "The robot made progress but did not fully complete the task"
      "Failure": "The robot failed to make meaningful progress on the task"

  - annotation_type: text
    name: action_description
    description: "Describe the sequence of actions the robot performed"

  - annotation_type: likert
    name: execution_quality
    description: "Rate the overall quality of the robot's execution"
    min_label: "Very Poor"
    max_label: "Excellent"
    size: 5

  - annotation_type: video_annotation
    name: action_phases
    description: "Segment the video into distinct action phases"
    mode: "segment"
    labels:
      - "Reaching"
      - "Grasping"
      - "Placing"
      - "Moving"
      - "Idle"

annotation_instructions: |
  You will be shown a task instruction and a video of a robot attempting to complete it.
  1. Read the task instruction carefully.
  2. Watch the full video of the robot's execution.
  3. Judge whether the task was a Success, Partial Success, or Failure.
  4. Describe the sequence of actions the robot performed.
  5. Rate the overall execution quality on a 5-point scale.
  6. Segment the video into action phases: Reaching, Grasping, Placing, Moving, or Idle.

html_layout: |
  <div style="padding: 15px; max-width: 800px; margin: auto;">
    <div style="background: #fefce8; border: 1px solid #fde68a; border-radius: 8px; padding: 16px; margin-bottom: 16px;">
      <strong style="color: #a16207;">Task Instruction:</strong>
      <p style="font-size: 16px; line-height: 1.6; margin: 8px 0 0 0;">{{text}}</p>
    </div>
    <div style="background: #1e293b; border-radius: 8px; padding: 16px; margin-bottom: 16px; text-align: center;">
      <video controls style="max-width: 100%; border-radius: 4px;">
        <source src="{{video_url}}" type="video/mp4">
        Your browser does not support the video tag.
      </video>
    </div>
  </div>

allow_all_users: true
instances_per_annotator: 50
annotation_per_instance: 2
allow_skip: true
skip_reason_required: false

Sample Datasample-data.json

json
[
  {
    "id": "rt2_001",
    "text": "Pick up the red apple from the table and place it in the bowl.",
    "video_url": "videos/robot_episode_001.mp4"
  },
  {
    "id": "rt2_002",
    "text": "Move the blue cup to the left side of the counter.",
    "video_url": "videos/robot_episode_002.mp4"
  }
]

// ... and 8 more items

Get This Design

View on GitHub

Clone or download from the repository

Quick start:

git clone https://github.com/davidjurgens/potato-showcase.git
cd potato-showcase/multimodal/rt2-robotic-action-annotation
potato start config.yaml

Dataset & paper

Brohan et al., CoRL 2023

Citation (BibTeX)

bibtex
@inproceedings{brohan2023rt2,
    title = "{RT}-2: Vision-Language-Action Models Transfer Web Knowledge to Robotic Control",
    author = "Brohan, Anthony and Brown, Noah and Carbajal, Justice and Chebotar, Yevgen and Chen, Xi and Choromanski, Krzysztof and Ding, Tianli and Driess, Danny and Dubey, Avinava and Finn, Chelsea and others",
    booktitle = "Proceedings of the Conference on Robot Learning (CoRL)",
    year = "2023",
    url = "https://arxiv.org/abs/2307.15818"
}

Details

Annotation Types

radiotextlikertvideo_annotation

Domain

RoboticsMultimodalEvaluation

Use Cases

Robotic ManipulationAction RecognitionTask Evaluation

Tags

roboticsrt2manipulationaction-segmentationcorl2023vision-language

Found an issue or want to improve this design?

Open an Issue