User registration in chromadb

Dockerfile

@@ -1,6 +1,11 @@
 # Use an official Python runtime as a parent image
 FROM python:3.9
 
+# Install system dependencies required by OpenCV
+RUN apt-get update && apt-get install -y \
+    libgl1-mesa-glx \
+ && rm -rf /var/lib/apt/lists/*
+
 # Create a non-root user with a specified user ID
 RUN useradd -m -u 1000 user
 
@@ -29,5 +34,9 @@ RUN pip install --no-cache-dir --user -r requirements.txt
 # Make port 7860 available to the world outside this container
 EXPOSE 7860
 
+# Indicate that a volume is expected at /home/user/data
+# This directory is intended for persistent storage
+VOLUME /home/user/data
+
 # Run the FastAPI application using Uvicorn, binding to port 7860
 CMD ["uvicorn", "app.main:app", "--host", "0.0.0.0", "--port", "7860"]

README.md

@@ -18,7 +18,6 @@ EduConnect/
 │   ├── dependencies.py              # Dependency utilities for JWT token verification, etc.
 │   ├── models.py                    # Database models for ORM
 │   ├── schemas.py                   # Pydantic schemas for request and response validation
-│   ├── crud.py                      # CRUD operations interfacing with the database
 │   ├── api/
 │   │   ├── __init__.py
 │   │   ├── userlogin.py             # Endpoint for user login functionality
@@ -33,8 +32,7 @@ EduConnect/
 │   │       └── user_registration.html # Template for user registration page
 │   └── utils/
 │       ├── __init__.py
-│       ├── authentication.py        # Integrates MTCNN and Facenet for login authentication
-│       └── database.py              # ChromaDB integration and database utilities
+│       └── ec_image_utils.py        # Integrates MTCNN and Facenet for login authentication
 ├── static/
 │   ├── css/
 │   ├── js/

app/admin/admin_functions.py

@@ -2,6 +2,52 @@ from fastapi import HTTPException, UploadFile, File, Form
 from typing import Optional
 import bcrypt
 import os
+from ..utils import get_user_cropped_image_from_photo
+
+
+
+# Registrering a face
+async def register_user(db, email: str, name: str, role: str, file: UploadFile = File(...)):
+    """
+    Processes and stores the image uploaded into vectordb as image embeddings.
+
+    :param db: The vector db collection handle to which the image embedding with email id as key will be upserted
+    :param email: The email id of the user being registered, this is assumed to be unique per user record
+    :param name: The user name (different from email) for display
+    :param role: The role associated with the user, it can only be student or teacher
+    :param file: The facial image of the user being registered, the first recognized face image would be used.
+
+    :return: email 
+    """
+    unique_filename = f"{email}.jpg"  # Use the email as the filename
+    file_path = f"/home/user/data/tmp/{unique_filename}"  # Specify your upload directory
+
+    # Ensure the directory exists
+    os.makedirs(os.path.dirname(file_path), exist_ok=True)
+
+    # Then, proceed to open the file
+    with open(file_path, "wb") as buffer:
+        contents = await file.read()
+        buffer.write(contents)
+
+    # Process the image to extract the face
+    cropped_face = get_user_cropped_image_from_photo(file_path)
+    
+    if cropped_face is not None:
+        
+        # Here you can store the embeddings along with user details in ChromaDB
+        # chroma_db.save_embeddings(user_id, embeddings)
+        db.upsert(images=[cropped_face], ids=[email], metadatas=[{"name":name, "role":role}])
+        return {"status": "User registered successfully", "image": cropped_face}
+
+    else:
+
+        return {"error": "No faces detected"}
+    
+    #os.remove(file_path)  # Optionally remove the file after processing, if not needed
+    
+
+
 
 # Admin Authentication
 def verify_admin_password(submitted_user: str, submitted_password: str) -> bool:
@@ -21,27 +67,6 @@ def verify_admin_password(submitted_user: str, submitted_password: str) -> bool:
 
     return False
 
-
-# User Registration
-async def register_user(email: str, name: str, role: str, file: UploadFile = File(...)) -> Optional[str]:
-    """
-    Registers a new user with the provided details and stores the profile picture.
-    
-    :param email: The user's email address.
-    :param name: The user's full name.
-    :param role: The user's role (e.g., Student, Teacher).
-    :param file: The profile picture file.
-    :return: User ID of the newly registered user or None if registration failed.
-    """
-    # Here, you would include logic to:
-    # 1. Process and validate the input data.
-    # 2. Use MTCNN and Facenet (or similar) to process the profile picture.
-    # 3. Store the user's details and the processed picture in ChromaDB.
-    # 4. Return the user ID or None if the registration fails.
-    
-    # This is a placeholder for the implementation.
-    pass
-
 # Additional Admin Functions
 # You could include other administrative functionalities here, such as:
 # - Listing all registered users.

app/admin/templates/admin_login.html

@@ -1,6 +1,6 @@
 <html>
   <head>
-    <title>EduConnect Administration- login page</title>
+    <title>EduConnect Administration - login page</title>
     <meta name="viewport" content="width=device-width, initial-scale=1.0">
     <link href="static/css/mvp.css" rel="stylesheet" media="screen">
   </head>

app/admin/templates/registration_success.html

@@ -0,0 +1 @@
+Success!

app/admin/templates/user_registration.html

@@ -1,11 +1,12 @@
 <!DOCTYPE html>
 <html>
 <head>
-    <title>User Registration</title>
+    <title>Econnect User Registration</title>
     <meta name="viewport" content="width=device-width, initial-scale=1.0">
     <link href="static/css/mvp.css" rel="stylesheet" media="screen">
 </head>
 <body>
+    <div class=""container">
     <h2>User Registration</h2>
     <form action="/admin/register_user" method="post" enctype="multipart/form-data">
         <label for="email">User ID (Email):</label>
@@ -23,10 +24,11 @@
         <label for="file">Profile Picture (.jpg):</label>
         <input type="file" id="file" name="file" accept=".jpg" required><br>
 
-        <button type="submit">Register</button>
+        <button type="submit" class="btn btn-default" >Register</button>
     </form>
     {% if error %}
         <p class="error"><strong>Error:</strong> {{ error }}
     {% endif %}
+    </div>
 </body>
 </html>

app/main.py

@@ -1,3 +1,5 @@
+import chromadb
+
 from fastapi import FastAPI, Request, Form, File, UploadFile, Depends
 from fastapi.middleware.cors import CORSMiddleware
 from fastapi.staticfiles import StaticFiles
@@ -5,9 +7,21 @@ from fastapi.responses import FileResponse, HTMLResponse, RedirectResponse
 from fastapi.templating import Jinja2Templates
 
 from .admin import admin_functions as admin
+from .utils.ec_image_utils import UserFaceEmbeddingFunction
 
 app = FastAPI()
 
+# Persitent storage for chromadb setup in /data volume
+ec_client = chromadb.PersistentClient("/home/user/data/chromadb")
+user_faces_db = ec_client.get_or_create_collection(name="user_faces_db", embedding_function=UserFaceEmbeddingFunction())
+
+@app.on_event("startup")
+async def startup_event():
+    # Perform any necessary ChromaDB setup or checks here
+    ec_client.heartbeat()
+    print("ChromaDB setup completed.")
+
+
 # Add middleware
 # Set all origins to wildcard for simplicity, but you should limit this in production
 app.add_middleware(
@@ -39,10 +53,16 @@ async def handle_admin_login(request: Request, username: str = Form(...), passwo
         # Reload login page with error message
         return templates.TemplateResponse("admin_login.html", {"request": request, "error": "Invalid password"})
 
+# To display the register user page
+@app.get("/admin/register_user", response_class=HTMLResponse)
+async def get_user_registration(request: Request):
+    # Render the registration form
+    return templates.TemplateResponse("user_registration.html", {"request": request})
+
 # User Registration Handler
 @app.post("/admin/register_user", response_class=HTMLResponse)
 async def handle_user_registration(request: Request, email: str = Form(...), name: str = Form(...), role: str = Form(...), file: UploadFile = File(...)):
-    user_id = await admin.register_user(email, name, role, file)
+    user_id = await admin.register_user(user_faces_db, email, name, role, file)
     if user_id:
         # Redirect or display a success message
         return templates.TemplateResponse("registration_success.html", {"request": request})

app/utils/__init__.py

@@ -0,0 +1,2 @@
+# In utils/__init__.py
+from .ec_image_utils import get_user_cropped_image_from_photo

app/utils/ec_image_utils.py

@@ -0,0 +1,105 @@
+import cv2
+from mtcnn.mtcnn import MTCNN
+import numpy as np
+from keras_facenet import FaceNet
+from chromadb.api.types import EmbeddingFunction, Embeddings, Image, Images
+from typing import List
+
+
+def load_image(filename):
+    """Load an image from a file path."""
+    img = cv2.imread(filename)
+    return cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
+
+
+def normalize(img):
+    """Normalize the given image array."""
+    mean, std = img.mean(), img.std()
+    return (img - mean) / std
+
+
+def detect_faces_with_mtcnn(img):
+    """Detect faces in an image using MTCNN."""
+    detector = MTCNN()
+    bounding_boxes = []
+    detected_faces = detector.detect_faces(img)
+    for detected_face in detected_faces:
+        bounding_boxes.append(detected_face["box"])
+    return bounding_boxes
+
+
+# Crops out parts of an image based on a list of bounding
+# boxes. The cropped faces are also resized to 160x160 in
+# preparation for passing it to FaceNet to compute the
+# face embeddings.
+#
+def crop_faces_to_160x160(img, bounding_boxes):
+    """Crop faces from an image based on detections."""
+    cropped_faces = []
+
+    for (x,y,w,h) in bounding_boxes:
+        cropped_face = img[y:y+h, x:x+w]
+        normalize(cropped_face)
+        cropped_face = cv2.resize(cropped_face, (160, 160), interpolation=cv2.INTER_CUBIC)
+        cropped_faces.append(cropped_face)
+
+    return np.array(cropped_faces)
+
+
+# Extract cropped user face image
+#
+def get_user_cropped_image_from_photo(filename):
+
+    # Load the image.
+    #...#
+    img = load_image(filename)
+
+
+    # Detect faces and extract all bounding boxes
+    #...#
+    bounding_boxes = detect_faces_with_mtcnn(img)
+
+
+    # Crop out the faces from the image
+    #...#
+    cropped_faces = crop_faces_to_160x160(img, bounding_boxes)
+
+    if cropped_faces.shape[0] == 0:
+        return
+
+
+    # Take the image of only the first detected face
+    #...#
+    cropped_face = cropped_faces[0:1, :, :, :]
+
+
+    # Get the face embeddings using FaceNet and return
+    # the results.
+    #...#
+    return cropped_face[0]
+
+
+###### Class implementing Custom Embedding function for chroma db
+#
+class UserFaceEmbeddingFunction(EmbeddingFunction[Images]):
+    def __init__(self):
+        # Intitialize the FaceNet model
+        self.facenet = FaceNet()
+
+    def __call__(self, input: Images) -> Embeddings:
+        # Since the input images are assumed to be `numpy.ndarray` objects already,
+        # we can directly use them for embeddings extraction without additional processing.
+        # Ensure the input images are pre-cropped face images ready for embedding extraction.
+
+        # Extract embeddings using FaceNet for the pre-cropped face images.
+        embeddings_array = self.facenet.embeddings(input)
+
+        # Convert numpy array of embeddings to list of lists, as expected by Chroma.
+        return embeddings_array.tolist()
+
+
+# Usage example:
+# user_face_embedding_function = UserFaceEmbeddingFunction()
+# Assuming `images` is a list of `numpy.ndarray` objects where each represents a pre-cropped face image ready for embedding extraction.
+# embeddings = user_face_embedding_function(images)
+

requirements.txt

@@ -2,13 +2,15 @@ fastapi==0.95.2                # Core framework for building APIs.
 uvicorn[standard]==0.18.3      # ASGI server for FastAPI, supports live reloading.
 requests==2.28.*               # For making HTTP requests, if needed by your app.
 torch==1.11.*                  # PyTorch, for handling deep learning models.
-transformers==4.*              # From Hugging Face, for working with pre-trained LLMs.
 sentencepiece==0.1.*           # For chat text processing
 mtcnn==0.1.1                   # For face detection in images.
 python-jose[cryptography]==3.3.*  # For creating, parsing, and verifying JWT tokens.
 python-multipart==0.0.5        # Necessary for form data handling, including file uploads.
-numpy                          # Fundamental package for scientific computing.
+numpy==1.23.0                  # Updated to resolve conflict
 chromadb==0.4.22               # Vector database interaction libraries.
 keras-facenet==0.3.2           # For face recognition and embedding, used alongside MTCNN.
 jinja2==3.0.*                  # For Admin site redndering
-bcrypt==4.1.*                  # For hashing secrets
+bcrypt==4.1.*                  # For hashing secrets
+opencv-python-headless==4.5.5.* # For image handling images
+tensorflow                     # Tensorflow is needed by MTCNN for facial recognition
+scipy                          # The scipy is required for keras-facenet