Robot learning data collection contractFrom designing collection conditions to data confirmation and delivery.
We are contracted to collect data necessary for imitation learning and operation verification of robots. We organize the target work, equipment, collection volume, and quality standards, and propose a process from test collection to actual collection and delivery.
Discuss data collection
Features/Contents that can be discussed
Design of conditions and quality standards
Define objects, work procedures, camera viewpoints, collection times, etc. Agree on success/failure judgments and exclusion criteria in advance.
From pilot collection to full-scale data collection
First, take a small sample and check what is recorded and how it will be incorporated into the learning environment. We will proceed with this collection after adjusting the conditions.
Confirm, organize and deliver
Check for missing or misaligned records and required metadata. We will organize and deliver the data in the agreed upon data format and folder structure.
Items to check before collection
Collection items and supported formats are determined by checking the signals that can be acquired by the target robot and the customer's learning environment.
- Target
- Robot, work, object, implementation environment
- data
- Collection items such as images, status, operation input, etc.
- Collection amount
- Combination of number of episodes, time, and conditions
- quality standards
- Success determination, loss, synchronization, exclusion conditions
- Delivery conditions
- Data format, scope of annotations, usage rights and storage period
Data acquisition by remote control
Discuss data collection
Please let us know the type of data you need, the intended learning use, the desired amount of collection, and the delivery date. If the specifications are not yet determined, you can consult with us regarding the scope of test collection. The cost will be estimated based on the work, quantity, location, etc.
Discuss data collection