Dataset Overview

Dataset Information

• Language: English
• License: CC BY‑NC‑ND 4.0
• Task Categories:
  • Image classification
  • Object detection
• Labels:
  • code

Dataset Details

• Features:
  • id (int32)
  • name (string)
  • image (image)
  • mask (image)
  • width (uint16)
  • height (uint16)
  • shapes (sequence) with sub‑features:
    • label (class_label) – label name Miner
    • type (string)
    • points (sequence of float32)
    • rotation (float32)
    • occluded (uint8)
    • attributes (sequence) containing:
      • name (string)
      • text (string)
• Split:
  • train: 8 samples, 5,907,438 bytes
• Download size: 5,795,853 bytes
• Dataset size: 5,907,438 bytes

Dataset Description

The dataset comprises photographs captured inside various mines, focusing on miners engaged in work. Each image includes bounding‑box detection of miners, and an attribute highlights whether the miner is sitting or standing.

The diverse applications—computer vision, safety assessment, etc.—make it a valuable resource for mining industry researchers, employers, and policy makers.

Dataset Structure

• images: Raw miner photographs
• boxes: Bounding‑box annotations for the raw images
• annotations.xml: XML files with bounding‑box coordinates and labels

Data Format

In the images folder, each picture is paired with an annotations.xml file that specifies the bounding‑box coordinates. For each point, an x and y coordinate is provided. The miner’s posture is also indicated via the is_sitting attribute (true/false).