README.md

metadata

title: Face Recognition Dataset (105 Classes)
emoji: 📸
colorFrom: yellow
colorTo: orange
sdk: python
tags:
  - dataset
  - face-recognition
  - embeddings
  - computer-vision
license: mit

📸 Face Recognition Dataset (105 Classes)

A curated and cleaned celebrity face dataset used for training and evaluating:

• Face Recognition Model (CNN Embeddings + SVM)
• Face Recognition Demo App (Streamlit)

This dataset contains 105 identities and ~18,000 manually organized images, formatted for deep-learning–based face recognition pipelines.

---

📁 Dataset Structure

The dataset follows a simple folder-based classification format:

face_recognition_dataset/
  ├── person_1/
  ├── person_2/
  ├── ...
  └── person_105/

Each folder contains multiple face images for that identity.
This structure is compatible with most ML frameworks and embedding-based models.

---

📦 Contents

• 18k+ images
• 105 celebrity identities
• Cleaned, resized, organized folder structure
• Suitable for:
  • Embedding extraction (FaceNet, ArcFace, etc.)
  • Classification (SVM, kNN, cosine similarity)
  • Clustering
  • Evaluation & benchmarking

---

🧠 Model Trained on This Dataset

The official model trained on this dataset is available at:

Model Repository: AI-Solutions-KK/face_recognition
Contains:

• svc_model.pkl
• classes.npy
• centroids.npy
• Metadata + reproducible training pipeline

The model achieves ~99% accuracy on this dataset.

---

🚀 Demo App Using This Dataset

A complete interactive app using this dataset is available at:

App Repository: AI-Solutions-KK/face_recognition_model_demo_app

Features:

• Image selection browser
• Real-time prediction
• Training report
• Prediction report
• Confusion matrix display

The app automatically downloads this dataset inside the Space using snapshot_download().

---

🧩 Recommended Usage

from huggingface_hub import snapshot_download

snapshot_download(
    repo_id="AI-Solutions-KK/face_recognition_dataset",
    repo_type="dataset",
    local_dir="my_dataset",
    local_dir_use_symlinks=False,
)

After download, the dataset will be available at:

my_dataset/face_recognition_dataset/<class>/<image>.jpg

---

🔧 Suitable For

• Face recognition research
• Deep metric learning
• Identity classification
• Transfer learning experiments
• Benchmarking models like:
  • FaceNet
  • ArcFace
  • MobileFaceNet
  • InsightFace

---

👤 Author

Developed and organized by Karan (AI-Solutions-KK)
Please ⭐ the repo if you find it useful.