Dataset Viewer
Auto-converted to Parquet Duplicate
image
imagewidth (px)
524
524
label
int64
0
8
label_name
stringclasses
9 values
8
Vegetation
4
Miscellaneous Trash
3
Metal
2
Glass
6
Plastic
5
Paper
2
Glass
3
Metal
4
Miscellaneous Trash
5
Paper
5
Paper
1
Food Organics
5
Paper
1
Food Organics
8
Vegetation
1
Food Organics
3
Metal
8
Vegetation
6
Plastic
6
Plastic
5
Paper
6
Plastic
8
Vegetation
2
Glass
5
Paper
6
Plastic
8
Vegetation
4
Miscellaneous Trash
3
Metal
2
Glass
1
Food Organics
7
Textile Trash
1
Food Organics
0
Cardboard
8
Vegetation
6
Plastic
0
Cardboard
1
Food Organics
8
Vegetation
2
Glass
3
Metal
3
Metal
6
Plastic
4
Miscellaneous Trash
6
Plastic
8
Vegetation
6
Plastic
2
Glass
8
Vegetation
4
Miscellaneous Trash
4
Miscellaneous Trash
6
Plastic
3
Metal
6
Plastic
4
Miscellaneous Trash
0
Cardboard
6
Plastic
5
Paper
8
Vegetation
3
Metal
7
Textile Trash
5
Paper
8
Vegetation
8
Vegetation
4
Miscellaneous Trash
8
Vegetation
5
Paper
0
Cardboard
2
Glass
3
Metal
6
Plastic
7
Textile Trash
7
Textile Trash
6
Plastic
8
Vegetation
6
Plastic
6
Plastic
3
Metal
3
Metal
5
Paper
0
Cardboard
5
Paper
3
Metal
6
Plastic
5
Paper
3
Metal
3
Metal
1
Food Organics
6
Plastic
3
Metal
6
Plastic
3
Metal
8
Vegetation
6
Plastic
6
Plastic
3
Metal
3
Metal
6
Plastic
3
Metal
6
Plastic
End of preview. Expand in Data Studio

Waste Segregation Dataset

Descripción

Este dataset contiene imágenes de residuos clasificadas en 9 categorías para tareas de segregación automática de desechos. Fue preparado como parte de un proyecto de aprendizaje automático en el Instituto Tecnológico de Costa Rica (ITCR), con el objetivo de entrenar y evaluar modelos de clasificación de imágenes basados en Vision Transformers (ViT).

Información del Dataset

  • Total de imágenes: ~4,765
  • División: 80% entrenamiento / 20% validación (del conjunto train original) + prueba independiente
  • Número de clases: 9
  • Formatos: JPEG, PNG, WebP
  • Fuente original: smarthkaushal/waste-segregation en Kaggle

Clases

Clase Descripción
Cardboard Cartón
Food Organics Residuos orgánicos de alimentos
Glass Vidrio
Metal Metal
Miscellaneous Trash Basura miscelánea
Paper Papel
Plastic Plástico
Textile Trash Residuos textiles
Vegetation Vegetación

Balance de Clases

El dataset presenta desbalance moderado: Plastic es la clase más representada (19%) y Textile Trash la menos representada (6.6%). La razón de desbalance (max/min) es ~2.89x.

Uso

from datasets import load_dataset
ds = load_dataset("ddompe/waste-segregation-dataset")

Licencia

CC BY 4.0 — Se permite uso, distribución y modificación con atribución.

Downloads last month
46