Used for
- Household robotic manipulationImitation learning researchRobotics demonstration collectionTask automation in domestic settingsDevelopment of scalable home robotics solutions
Features
- Open-source frameworkHousehold task learning109 tasks tested81% success rate20-minute task learningHoNY datasetPre-trained ResNet-34 model3D printable demonstration toolRGB-D video dataMIT licensed codeOpen-source framework for household robotic manipulationSupports 109 different tasksTested in 10 homes across New York City81% success rate in task executionLearns new tasks in 20 minutesUtilizes imitation learning techniquesIncludes a demonstration collection tool, 'The Stick'Compatible with 3D printed parts and iPhone for demonstrationOffers dataset creation code for further researchIncludes a GitHub repository for community collaboration
What is Dobb-E?
Dobb-E is an open-source framework for learning household robotic manipulation, allowing robots to perform tasks within home environments.How many tasks can Dobb-E perform?
Dobb-E supports 109 different tasks.Where has Dobb-E been tested?
Dobb-E has been tested in 10 homes across New York City.What is the success rate of Dobb-E?
Dobb-E has an 81% success rate in task execution.How long does it take for Dobb-E to learn a new task?
Dobb-E can learn a new task in just 20 minutes.What techniques does Dobb-E use for learning?
Dobb-E utilizes imitation learning techniques.What is 'The Stick'?
'The Stick' is a demonstration collection tool built from a Reacher-grabber stick, 3D printed parts, and an iPhone, used to collect demonstrations for robots.Is there a community or repository for Dobb-E?
Yes, there is a GitHub repository for community collaboration and further development.Can Dobb-E be used for research purposes?
Yes, Dobb-E includes dataset creation code and is suitable for imitation learning research.What are the applications of Dobb-E?
Dobb-E is used for household robotic manipulation, task automation in domestic settings, and the development of scalable home robotics solutions.