Pick-and-Place
Item picking from shelves, bins, and totes across thousands of SKU variations. Data includes grasp type, object category, weight class, and placement accuracy.
Warehouse robotics data collection
Warehouse Robotics Data Collection
Real-world warehouse robotics data for pick-and-place, palletizing, inventory scanning, and package handling. Multi-sensor demonstration datasets collected in operating logistics facilities.
On-site capture · warehouse floor
Why real-world data
Why Warehouse Robotics Needs Real-World Data
Warehouses are dynamic, high-throughput environments where conditions change shift to shift. SKU variation runs into the thousands — different sizes, weights, textures, fragility levels. Lighting changes as doors open, shelves get restocked, and seasonal inventory shifts the layout. A robot trained on a fixed set of objects in a controlled lab will fail when it encounters a new packaging format or a partially occluded item on a cluttered shelf.
Real-world warehouse data captures this variation naturally. Human operators working actual pick-and-place tasks encounter edge cases — unusual object shapes, damaged packaging, unexpected obstacles — that simulation cannot predict. Training on this data produces policies that handle the long tail of warehouse operations.
Storage racking · restock
Facility floor · live shift
Tasks
Warehouse Tasks We Collect Data For
Palletizing
Layer patterns, box stacking sequences, weight distribution optimization. Data captures placement coordinates, stack stability labels, and pallet configuration.
Inventory Scanning
Shelf scanning, barcode/label reading, stock level estimation. Data combines visual observations with location annotations and count verification.
Package Handling
Sorting, routing, and loading packages of varying sizes and fragility. Data includes manipulation trajectories, damage risk labels, and throughput metrics.
Tote & Bin Management
Filling, emptying, and organizing storage containers. Data captures packing strategies, space utilization, and item arrangement patterns.
Inter-Station Transport
Moving items between workstations, loading docks, and storage areas. Data includes navigation trajectories, obstacle avoidance, and timing coordination.
Data modalities
Data Modalities for Warehouse Robotics
Egocentric RGB-D from the operator's perspective, 6-DoF object poses, gripper state recordings, barcode/label recognition data, and spatial mapping of facility layout. All synchronized, timestamped, and calibrated.
Annotations include object identity, grasp type, placement accuracy, task completion status, and episode-level success/failure flags. Every data stream is aligned to a common clock so downstream models can fuse vision, depth, and action signals without additional preprocessing.
Egocentric RGB-D6-DoF object posesGripper stateBarcode / label recognitionSpatial mappingGrasp typeSuccess / failure flags
Egocentric RGB · hand tracking
Head-mounted stereo RGB with 21-keypoint hand skeletons, 30 Hz.
Depth · same clock
Lossless 16-bit depth, frame-paired with the RGB stream.
23
Channels on one clock per recording
30Hz
RGB, depth and action labels
21
3D keypoints per hand
3
Cameras per capture rig
Data Explorer
Inspect Warehouse Data in the Explorer
Collected warehouse robotics data is browsable through Humaid's data explorer. Teams can review individual pick-and-place sequences, verify object detection annotations across SKU variations, and inspect synchronized egocentric video with hand pose overlays — directly in the browser.
Each sequence exposes 60+ metadata properties, downloadable sensor files, and annotation overlays. This makes it straightforward to validate data quality before using it for behavior cloning or diffusion policy training. Open the data explorer (opens in a new tab).
Sequence view · synchronized streams
Get started
Start Collecting Warehouse Data
Tell us about your warehouse tasks, robot platform, and facility. We will design a capture protocol, deploy trained operators on-site, and deliver calibrated datasets ready for your training pipeline.
Related topics