imagenet1k_classes / README.md
mlnomad's picture
Upload README.md with huggingface_hub
d0d92d6 verified
---
license: mit
---
# ImageNet-1k Individual Class Datasets Hub
This dataset serves as a central hub and index for the **1,000 individual class datasets** derived from the original `imagenet-1k` dataset. Each class has been separated into its own repository for easy access and analysis.
This repository does not contain any images itself. Instead, it provides a `class_mapping.csv` file that maps each class ID and name to its corresponding dataset repository on the Hugging Face Hub.
## How to Use
The intended workflow is to use this repository to find the `repo_id` for a class you are interested in, and then load that specific dataset.
### 1. Load the Class Mapping File
First, load the `class_mapping.csv` file from this repository. You can do this directly using `pandas`.
```python
import pandas as pd
from datasets import load_dataset
# Load the mapping file from the Hub
mapping_csv_url = "[https://huggingface.co/datasets/](https://huggingface.co/datasets/)mlnomad/imagenet1k_classes/resolve/main/class_mapping.csv"
class_df = pd.read_csv(mapping_csv_url)
print("Class mapping loaded. Here's a preview:")
print(class_df.head())
```
### 2. Find and Load a Specific Class Dataset
Once you have the `class_df` DataFrame, you can easily find the repository for any class and load it using `datasets.load_dataset`.
#### Example: Loading the 'goldfish' class
```python
# Find the entry for 'goldfish'
# The class names can be long, so we use .str.contains() for a flexible search
goldfish_entry = class_df[class_df['class_name'].str.contains("goldfish", case=False)]
if not goldfish_entry.empty:
# Get the repo_id from the DataFrame
goldfish_repo_id = goldfish_entry.iloc[0]['repo_id']
print(f"Found 'goldfish' dataset at: {goldfish_repo_id}")
# Load the actual dataset containing the images
print("Loading dataset...")
goldfish_dataset = load_dataset(goldfish_repo_id)
print("\nSuccessfully loaded!")
print(goldfish_dataset)
# You can now access the images
# example_image = goldfish_dataset['train'][0]['image']
# example_image.show()
else:
print("Could not find the 'goldfish' class.")
```
## Class Mapping Preview (`class_mapping.csv`)
Here are the first 10 entries in the mapping file:
```
| class_id | class_name | repo_id |
|-----------:|:------------------------------------------------------------------------------------|:------------------------------------------------------------------------------------------------|
| 0 | tench, Tinca tinca | mlnomad/imnet1k_tench_Tinca_tinca |
| 1 | goldfish, Carassius auratus | mlnomad/imnet1k_goldfish_Carassius_auratus |
| 2 | great white shark, white shark, man-eater, man-eating shark, Carcharodon carcharias | mlnomad/imnet1k_great_white_shark_white_shark_man-eater_man-eating_shark_Carcharodon_carcharias |
| 3 | tiger shark, Galeocerdo cuvieri | mlnomad/imnet1k_tiger_shark_Galeocerdo_cuvieri |
| 4 | hammerhead, hammerhead shark | mlnomad/imnet1k_hammerhead_hammerhead_shark |
| 5 | electric ray, crampfish, numbfish, torpedo | mlnomad/imnet1k_electric_ray_crampfish_numbfish_torpedo |
| 6 | stingray | mlnomad/imnet1k_stingray |
| 7 | cock | mlnomad/imnet1k_cock |
| 8 | hen | mlnomad/imnet1k_hen |
| 9 | ostrich, Struthio camelus | mlnomad/imnet1k_ostrich_Struthio_camelus |
```