Datasets:
File size: 1,939 Bytes
a6b6e3d | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 | ---
license: other
task_categories:
- image-classification
- zero-shot-image-classification
tags:
- face-recognition
- face-verification
- lfw
- imagefolder
pretty_name: LFW HF-ready
size_categories:
- 10K<n<100K
---
# LFW HF-ready
This folder packages the local LFW (Labeled Faces in the Wild) images as a
Hugging Face `imagefolder` dataset with the canonical 10-fold verification
pairs file.
## Layout
```
lfw/
├── README.md
├── pairs.csv
└── train/
├── images/<shard>/<file>.jpg
└── metadata.csv
```
## metadata.csv columns
- `file_name`: relative image path used by `ImageFolder`, e.g. `images/000/Aaron_Eckhart_0001.jpg`.
- `label`: numeric identity label.
- `label_name` / `identity`: identity name.
- `image_num`: per-identity image index from the original filename (1-based).
- `source_filename`: original LFW filename.
## pairs.csv columns
`pairs.csv` mirrors the official LFW `pairs.txt` (10 folds x 300 positive +
300 negative = 6000 verification pairs).
- `pair_id` (0..5999), `fold_id` (1..10), `fold_position` (0..299).
- `is_same`: 1 for positive pairs (same identity), 0 for negatives.
- `image_a`, `image_b`: bare filenames (e.g. `Abel_Pacheco_0001.jpg`).
- `image_a_path`, `image_b_path`: paths under the `train` split.
## Local Stats
- Images: 13233
- Unique identities: 5749
- Identities with one image: 4069
- Verification pairs: 6000 (3000 positive / 3000 negative)
- Folds: 10 x 600 pairs
## Loading
```python
from datasets import load_dataset
ds = load_dataset("imagefolder", data_dir="data/evaluation/huggingface/lfw")
train = ds["train"]
```
```python
import pandas as pd
pairs = pd.read_csv("data/evaluation/huggingface/lfw/pairs.csv")
```
## Notes
LFW is described by its authors as an unconstrained face verification benchmark.
The images here are the original (non-aligned) drop. Check the original dataset
terms before publishing or redistributing it.
|