Gelsight-Welds: Paired Tactile and RGB Images of Weld Beads

Paired tactile and top-down RGB captures of weld beads on a flat sample plate. Each session includes a 10×10 grid of GelSight Mini tactile touches and corresponding overhead RGB images from an Intel RealSense D555 camera, cropped and aligned per tile. The dataset was created by author ppak10 and last updated on Hugging Face in June 2026.

Use Cases

• Train multimodal models for weld defect detection based on paired tactile and visual data.
• Develop sensor fusion algorithms for robotic inspection based on aligned tactile and RGB image pairs.
• Benchmark computer vision models for material surface analysis based on the described GelSight and RealSense data.
• Research robotic tactile perception for manufacturing based on the grid of GelSight touches.

Strengths

• Data includes paired captures from two distinct sensor types (GelSight Mini tactile and Intel RealSense D555 RGB).
• Each sample session is systematically scanned with a 10×10 grid of tactile touches.
• Overhead RGB images are cropped into per-tile images aligned with each tactile touch.

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

huggingface

Collection Method

Scanned with a 10×10 grid of GelSight Mini touches and an overhead Intel RealSense D555 camera.

Freshness

Last updated 2026-06-14 22:31:03; freshness should be verified.