knn.py

import numpy as np
import cv2
import base64
from tensorflow.keras.preprocessing import image
import logging
import joblib

# Load KNN model globally
knn_model = joblib.load('k-nn_model.joblib')

# Emotion classes
class_labels = ['angry', 'disgust', 'fear', 'happy', 'neutral', 'sad', 'surprise']

def detect_knn(image_path):
    frame = cv2.imread(image_path)
    gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
    resized = cv2.resize(gray, (48, 48))  # 48x48
    norm_img = resized / 255.0

    # Feature extraction similar to training: horizontal chunks
    chunks = [norm_img[:, i*8:(i+1)*8] for i in range(6)]  # 6 chunks of 8px
    sequence = np.stack([chunk.flatten() for chunk in chunks])  # (6, 384)
    features = sequence.flatten()  # (2304,)
    features = features[:994]  # match training shape

    features = features.reshape(1, -1)

    prediction = knn_model.predict(features)[0]
    class_labels = ['angry', 'disgust', 'fear', 'happy', 'neutral', 'sad', 'surprise']
    emotion = class_labels[prediction]

    _, buffer = cv2.imencode('.jpg', frame)
    frame_base64 = base64.b64encode(buffer).decode('utf-8')

    return emotion, frame_base64