Take a video and tear it down into frames that have faces then classify the person's emotions

process.py

import cv2
import os
import glob
from deepface import DeepFace

videos_folder = 'videos/'
output_folder = 'output_frames/'

if not os.path.exists(output_folder):
    os.makedirs(output_folder)

video_files = glob.glob(os.path.join(videos_folder, '*.webm'))
# This only reads .webm files because that's what the download youtube video gave me

face_cascade = cv2.CascadeClassifier(cv2.data.haarcascades + 'haarcascade_frontalface_default.xml')

for video_path in video_files:
    video_name = os.path.splitext(os.path.basename(video_path))[0]
    cap = cv2.VideoCapture(video_path)
    frame_rate = cap.get(cv2.CAP_PROP_FPS)

    frame_interval = 30  
    frame_number = 0

    success, frame = cap.read()
    while success:
        frame_number += 1

        if frame_number % frame_interval == 0:
            gray_frame = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
            faces = face_cascade.detectMultiScale(gray_frame, scaleFactor=1.1, minNeighbors=5)
            if len(faces) > 0:                
                (x, y, w, h) = faces[0]
                face_img = frame[y:y+h, x:x+w]
                
                try:
                    analysis = DeepFace.analyze(face_img, actions=['emotion'], enforce_detection=False)
                    
                    # Check if analysis is a list
                    if isinstance(analysis, list):
                        # Access the first element
                        analysis = analysis[0]
                    
                    emotion = analysis['dominant_emotion']
                except Exception as e:
                    print(f"Emotion recognition failed at frame {frame_number}: {e}")
                    emotion = 'unknown'

                
                timestamp = frame_number / frame_rate
                timestamp_formatted = "{:.2f}".format(timestamp)
                
                output_filename = f"{video_name}_{emotion}_{timestamp_formatted}.png"
                output_path = os.path.join(output_folder, output_filename)
                
                cv2.imwrite(output_path, frame)
                print(f"Saved frame at {timestamp_formatted}s to {output_filename}")

        success, frame = cap.read()

    cap.release()