CoolFace
Datasetpublic

iDharshan/metal_scrap_dataset

Metal Scrap Dataset Overview This dataset contains synthetic images of metal scraps, categorized into three grades based on quality and recyclability. The images were generated using Stable Diffusion and augmented to enhance variety and robustness for machine learning tasks. Dataset Details Total Images: ~13,895 (after augmentation) Categories: Grade A: 5,085 images (1,017 original) - Clean, rust-free metals. Grade B: 5,570 images (1,114 original)… See the full description on the dataset page: https://huggingface.co/datasets/iDharshan/metal_scrap_dataset.

sourceHugging Faceupdated 2y agoView on Hugging Face
1likes8downloads
Dataset Card

Metal Scrap Dataset

Overview

This dataset contains synthetic images of metal scraps, categorized into three grades based on quality and recyclability. The images were generated using Stable Diffusion and augmented to enhance variety and robustness for machine learning tasks.

Dataset Details

  • —Total Images: ~13,895 (after augmentation)
  • —Categories:
  • —Grade A: 5,085 images (1,017 original) - Clean, rust-free metals.
  • —Grade B: 5,570 images (1,114 original) - Lightly rusted, recyclable metals.
  • —Grade C: 3,240 images (648 original) - Unrecyclable, heavily degraded scrap.
  • —Image Size: 224x224 pixels
  • —Format: PNG

Data Augmentation

Each original image was resized and augmented with 4 variations using the following transforms:

  • —Resize to 224x224
  • —Random horizontal and vertical flips
  • —Random rotation (up to 30°)
  • —Color jitter (brightness, contrast, saturation, hue)
  • —Random affine (rotation, translation, scaling)
  • —Random perspective distortion
  • —Random grayscale conversion

Usage

Intended for training models in metal scrap classification, recycling analysis, or computer vision research. Load the dataset via Hugging Face:

python
from datasets import load_dataset
dataset = load_dataset("iDharshan/metal_scrap_dataset")