RCS UTN Green Box: Multi-View Robot Manipulation Videos with Proprioception

University of Technology Nuremberg collected this grouped video dataset of a multi-view robot manipulation task using the Robot Control Stack ecosystem. Each episode contains synchronized videos from multiple cameras, along with dense robot proprioception and action data on every frame. The dataset was published by Voxel51 and last updated on June 8, 2026.

Use Cases

• Training robot manipulation policies based on synchronized multi-view video streams.
• Developing models for action recognition in robotic tasks based on dense proprioception data.
• Benchmarking multi-view video understanding algorithms for robotics applications.
• Studying robot proprioception and action sequences for imitation learning.

Strengths

• Data includes synchronized multi-view video streams for each episode.
• Each frame contains dense robot proprioception and action data.
• Dataset is structured as a grouped FiftyOne dataset, which suggests organized multi-modal data.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

University of Technology Nuremberg, Robot Control Stack (RCS) ecosystem.

Collection Method

Collected as part of a robot manipulation task ('pick the green box').

Freshness

Last updated 2026-06-08 16:49:26; freshness should be verified.