README.md

metadata

license: apache-2.0
task_categories:
  - image-classification
tags:
  - imagenet
  - computer-vision
  - imagenet-100
  - image-compression
  - decolorization
size_categories:
  - 100K<n<1M

ImageNet-100 Dataset (Zipped ImageFolder)

Overview

This dataset is a 100-class subset of the ImageNet-2012 (ILSVRC2012) dataset. It was specifically curated for academic research in computer vision, including tasks such as image compression and color-to-grayscale conversion (decolorization).

Dataset Details

• Class Selection: 100 classes were selected using a fixed random seed (42) to ensure reproducibility.
• Format: The dataset is stored as zipped chunks (10 files total) to facilitate stable uploads and high-speed downloads.
• Total Splits:
  • Train: ~130,000 images, distributed across 5 ZIP chunks (part1 to part5).
  • Validation: 5,000 images, distributed across 5 ZIP chunks (part1 to part5).
• Labels: A complete mapping of WordNet IDs to human-readable labels is included in the Labels.json file.

📂 Class Categories

The 100 selected classes cover a diverse range of categories:

Category	Count	Examples
Canines (Dogs)	13	Siberian husky, Bloodhound, Miniature schnauzer
Birds	8	Hummingbird, Sulphur-crested cockatoo, Goose
Primates	4	Chimpanzee, Howler monkey, Macaque
Wild Mammals	12	Polar bear, Hippopotamus, Red panda, White wolf
Reptiles & Fish	8	Stingray, Bullfrog, Alligator lizard
Vehicles	5	Minibus, Moped, Trailer truck
Instruments	3	Flute, Bassoon, Trombone
Household Items	18	Teapot, Hourglass, Vacuum cleaner, Reflex camera
Food & Nature	7	Banana, Mushroom, Seashore, Potpie
Sports & Other	22	Volleyball, Baseball, Scuba diver, Stone wall

---

How to Use

1. Automatic Download and Extraction

Since the data is split into multiple chunks, use the following script to reconstruct the ImageFolder structure.

import os
import zipfile
from huggingface_hub import hf_hub_download

repo_id = "asafaa/imagent100"
target_dir = "./imagenet100"

def download_and_extract(split, num_parts=5):
    for i in range(1, num_parts + 1):
        filename = f"imagenet100_{split}_part{i}.zip"
        print(f"Downloading {filename}...")
        path = hf_hub_download(repo_id=repo_id, filename=filename, repo_type="dataset")
        
        with zipfile.ZipFile(path, 'r') as zip_ref:
            zip_ref.extractall(target_dir)
    print(f"Finished extracting {split} split.")

# Download both splits
download_and_extract("train")
download_and_extract("val")