Update README.md

README.md

@@ -1,6 +1,6 @@
 ---
 license: mit
-language:
+language: 
   - en
 metrics:
   - accuracy
@@ -14,35 +14,36 @@ tags:
   - facenet
   - computer_vision
   - streamlit
-library_name: custom
-base_model: custom
+  - cpu_friendly
 datasets:
   - AI-Solutions-KK/face_recognition_demo_dataset
 ---
 
-# Face Recognition Model_(CNN embeddings + SVM)
-DEEP LEARNING & ML JOINTLY USED TO SAVE COMPUTATIONAL POWER 
+# 🧠 Face Recognition Model (CNN Embeddings + SVM)
+### **Deep Learning + Machine Learning Combined for Efficient CPU-Based Face Recognition**
 
-This repository stores my trained face-recognition model.  
-It contains the SVM classifier and supporting numpy files used in my
-**Face Recognition System** (Streamlit demo).
+This repository stores my trained **Face Recognition Model** using:
 
-The model is trained on FaceNet embeddings and is designed to run
-efficiently on CPU.
+- **FaceNet (InceptionResnetV1)** to extract 512-D face embeddings  
+- **SVM Classifier** for identity recognition  
+- **Centroid baseline** for fast cosine-similarity checks  
 
----
+Built to run **efficiently on CPU**, making it ideal for lightweight deployment, low-power systems, and Streamlit apps.
 
-## 🧩 Files in this repo
+---
 
-- `svc_model_retrained.pkl` – SVM classifier trained on FaceNet embeddings  
-- `centroids.npy` – class centroids for cosine-similarity baseline  
-- `classes.npy` – list of class labels (one per identity)
+## 📁 Files in this Repository
 
-(If your file names are slightly different, keep the same idea but change the names.)
+| File | Description |
+|------|-------------|
+| `svc_model_retrained.pkl` | SVM classifier trained on FaceNet embeddings |
+| `centroids.npy` | Class centroids for cosine-similarity baseline |
+| `classes.npy` | List of all identity labels |
+| `README.md` | This model card |
 
 ---
 
-## 🚀 How to load this model in Python
+# 🚀 How to Load This Model in Python
 
 ```python
 from huggingface_hub import hf_hub_download
@@ -59,3 +60,91 @@ classes_path = hf_hub_download(REPO_ID, "classes.npy")
 svc_model = joblib.load(svc_path)
 centroids = np.load(centroids_path)
 class_names = np.load(classes_path, allow_pickle=True)
+
+print("Model loaded successfully!")
+```
+
+---
+
+# 🔮 Simple Inference Example (Predict Face Identity)
+
+```python
+from huggingface_hub import hf_hub_download
+import joblib
+import numpy as np
+from PIL import Image
+import torch
+from facenet_pytorch import InceptionResnetV1
+import cv2
+
+REPO_ID = "AI-Solutions-KK/face_recognition"
+
+# Download model
+svc_path = hf_hub_download(REPO_ID, "svc_model_retrained.pkl")
+centroids_path = hf_hub_download(REPO_ID, "centroids.npy")
+classes_path = hf_hub_download(REPO_ID, "classes.npy")
+
+svc_model = joblib.load(svc_path)
+centroids = np.load(centroids_path)
+class_names = np.load(classes_path, allow_pickle=True)
+
+# Load FaceNet backbone
+facenet = InceptionResnetV1(pretrained="vggface2").eval()
+
+def preprocess(img_path):
+    img = Image.open(img_path).convert("RGB")
+    img = np.array(img)
+    img = cv2.resize(img, (160, 160))
+    img = img.astype("float32") / 255.0
+    img = torch.tensor(img).permute(2, 0, 1).unsqueeze(0)
+    return img
+
+def get_embedding(img_path):
+    img = preprocess(img_path)
+    with torch.no_grad():
+        emb = facenet(img).numpy()
+    return emb
+
+def predict_face(img_path):
+    emb = get_embedding(img_path)
+    pred = svc_model.predict(emb)[0]
+    confidence = np.max(svc_model.decision_function(emb))
+    return pred, confidence
+
+# -------- RUN ----------
+img_path = "test.jpg"
+label, prob = predict_face(img_path)
+
+print("Predicted Identity:", label)
+print("Confidence Score:", prob)
+```
+
+---
+
+# 🧑‍🔧 For Developers — Train on Your Own Dataset  
+This model is intended as a **plug-and-play template**.
+
+Just replace the dataset with your own and retrain:
+
+- Extract FaceNet embeddings  
+- Train SVM  
+- Upload 3 files:
+  - `svc_model.pkl`
+  - `centroids.npy`
+  - `classes.npy`
+
+You're done.
+
+---
+
+If you want, I can also prepare a **professional HF Model Card** with:  
+✔ Model Architecture  
+✔ Training Procedure  
+✔ Evaluation Metrics  
+✔ Limitations  
+✔ Intended Use / Misuse  
+✔ Citations  
+
+Just say **“make model card pro version”**.
+
+Let me know when to update!