Upload 3 files

app.py

@@ -0,0 +1,174 @@
+import os
+import sys
+import traceback
+import gradio as gr
+import cv2 as cv
+import numpy as np
+import mediapipe as mp
+from utils import blinkRatio
+
+def custom_excepthook(type, value, tb):
+    traceback.print_exception(type, value, tb)
+    sys.__excepthook__(type, value, tb)
+
+sys.excepthook = custom_excepthook
+
+def list_overlay_images(directory):
+    return [f for f in os.listdir(directory) if f.endswith('.png')]
+
+def process_frame(frame, overlay, LEFT_EYE, RIGHT_EYE, LEFT_IRIS, RIGHT_IRIS, 
+                  min_detection_confidence, min_tracking_confidence, alpha):
+    try:
+        mp_face_mesh = mp.solutions.face_mesh
+        with mp_face_mesh.FaceMesh(
+            max_num_faces=1,
+            refine_landmarks=True,
+            min_detection_confidence=min_detection_confidence,
+            min_tracking_confidence=min_tracking_confidence
+        ) as face_mesh:
+            rgb_frame = cv.cvtColor(frame, cv.COLOR_BGR2RGB)
+            rgba_frame = cv.cvtColor(frame, cv.COLOR_BGR2RGBA)
+            height, width = rgba_frame.shape[:2]
+            results = face_mesh.process(rgb_frame)
+            if results.multi_face_landmarks:
+                zero_overlay = np.zeros_like(rgba_frame)
+                mesh_points = np.array([np.multiply([p.x, p.y],
+                        [width, height]).astype(int) for p in results.multi_face_landmarks[0].landmark])
+                iris_mask_left = np.zeros(rgba_frame.shape, dtype=np.uint8)
+                iris_mask_right = np.zeros(rgba_frame.shape, dtype=np.uint8)
+                _, re_ratio, le_ratio = blinkRatio(rgb_frame, mesh_points, RIGHT_EYE, LEFT_EYE)
+                (l_cx, l_cy), l_radius = cv.minEnclosingCircle(mesh_points[LEFT_IRIS])
+                (r_cx, r_cy), r_radius = cv.minEnclosingCircle(mesh_points[RIGHT_IRIS])
+                center_left = (int(l_cx), int(l_cy))
+                center_right = (int(r_cx), int(r_cy))
+                cv.circle(iris_mask_left, center_left, int(l_radius), (255, 0, 0, 255), -1, cv.LINE_AA)
+                cv.circle(iris_mask_right, center_right, int(r_radius), (255, 0, 0, 255), -1, cv.LINE_AA)
+                bbx_size_l = int((l_radius * 2) / 2)
+                bbx_size_r = int((r_radius * 2) / 2)
+                resized_overlay_l = cv.resize(overlay, (bbx_size_l * 2, bbx_size_l * 2), interpolation=cv.INTER_CUBIC)
+                resized_overlay_r = cv.resize(overlay, (bbx_size_r * 2, bbx_size_r * 2), interpolation=cv.INTER_CUBIC)
+                y1_r = center_right[1] - bbx_size_r
+                y2_r = center_right[1] + bbx_size_r
+                x1_r = center_right[0] - bbx_size_r
+                x2_r = center_right[0] + bbx_size_r
+                y1_l = center_left[1] - bbx_size_l
+                y2_l = center_left[1] + bbx_size_l
+                x1_l = center_left[0] - bbx_size_l
+                x2_l = center_left[0] + bbx_size_l
+                if (resized_overlay_l.shape == zero_overlay[y1_l:y2_l, x1_l:x2_l].shape) & (le_ratio < 5.0) & (le_ratio > 2.0):
+                    zero_overlay[y1_l:y2_l, x1_l:x2_l] = resized_overlay_l
+                if (resized_overlay_r.shape == zero_overlay[y1_r:y2_r, x1_r:x2_r].shape) & (re_ratio < 5.0) & (re_ratio > 2.0):
+                    zero_overlay[y1_r:y2_r, x1_r:x2_r] = resized_overlay_r
+                eye_mask_left = np.zeros(rgba_frame.shape, dtype=np.uint8)
+                eye_mask_right = np.zeros(rgba_frame.shape, dtype=np.uint8)
+                cv.fillPoly(eye_mask_left, [mesh_points[LEFT_EYE]], (255, 0, 0, 255))
+                cv.fillPoly(eye_mask_right, [mesh_points[RIGHT_EYE]], (255, 0, 0, 255))
+                zero_overlay[np.where((iris_mask_left[:, :, 3] > 0) & (eye_mask_left[:, :, 3] == 0))] = 0
+                zero_overlay[np.where((iris_mask_right[:, :, 3] > 0) & (eye_mask_right[:, :, 3] == 0))] = 0
+                rgba_frame = cv.addWeighted(rgba_frame, 1, zero_overlay, alpha, 0)
+        return rgba_frame
+    except Exception as e:
+        print(f"Error in process_frame: {e}")
+        traceback.print_exc()
+
+def process_image(input_image, overlay_file, alpha=0.3, min_detection_confidence=0.5, min_tracking_confidence=0.5):
+    overlay_file = overlay_file + '.png'
+    overlay_path = os.path.join(os.getcwd(),'overlays', overlay_file)
+    overlay = cv.imread(overlay_path, cv.IMREAD_UNCHANGED)
+    frame = np.array(input_image)
+    processed_frame = process_frame(frame, overlay, LEFT_EYE, RIGHT_EYE, LEFT_IRIS, RIGHT_IRIS, 
+                                    min_detection_confidence, min_tracking_confidence, alpha)
+    return cv.cvtColor(processed_frame, cv.COLOR_BGR2RGB)
+
+def process_video(input_video, overlay_file, alpha=0.3, output_format='mp4', output_frame_rate=30,
+                                    min_detection_confidence=0.5, min_tracking_confidence=0.5):
+    overlay_file = overlay_file + '.png'
+    overlay_path = os.path.join(os.getcwd(),'overlays', overlay_file)
+    overlay = cv.imread(overlay_path, cv.IMREAD_UNCHANGED)
+    cap = cv.VideoCapture(input_video)
+    output_path = os.path.join(os.getcwd(),f'video_processed.{output_format}')
+    # Define the codec and create a VideoWriter object to save the processed video
+    if (not isinstance(overlay,type(None))) & (not isinstance(cap,type(None))):
+        # Get the dimensions of the frame, fps
+        fps=int(output_frame_rate)
+        if fps==0:
+            fps = cap.get(5)
+        ret, frame = cap.read()
+        height, width, _ = frame.shape
+        fourcc = cv.VideoWriter_fourcc(*'mp4v' if output_format == 'mp4' else 'MJPG')
+        out = cv.VideoWriter(output_path, fourcc, fps, (width, height))  
+        while(cap.isOpened()):
+            ret, frame = cap.read()
+            if ret == True:
+                processed_frame = process_frame(frame,overlay,LEFT_EYE, RIGHT_EYE, LEFT_IRIS, RIGHT_IRIS, 
+                                                float(min_detection_confidence), 
+                                                float(min_tracking_confidence), float(alpha))  # Assuming process_frame is a function that processes a single frame
+                processed_frame = cv.cvtColor(processed_frame, cv.COLOR_RGBA2BGR)
+                out.write(processed_frame)
+            else:
+                break
+        cap.release()
+        out.release()
+    return output_path
+
+
+def process_webcam(frame, overlay_file, alpha=0.3, min_detection_confidence=0.5, min_tracking_confidence=0.5):
+    overlay_file = overlay_file + '.png'
+    overlay_path = os.path.join(os.getcwd(), overlay_file)
+    overlay = cv.imread(overlay_path, cv.IMREAD_UNCHANGED)
+    processed_frame = process_frame(frame, overlay, LEFT_EYE, RIGHT_EYE, LEFT_IRIS, RIGHT_IRIS, 
+                                    min_detection_confidence, min_tracking_confidence, alpha)
+    return processed_frame
+
+LEFT_EYE = [362, 382, 381, 380, 374, 373, 390, 249, 263, 466, 388, 387, 386, 385, 384, 398]
+RIGHT_EYE = [33, 7, 163, 144, 145, 153, 154, 155, 133, 173, 157, 158, 159, 160, 161, 246]
+LEFT_IRIS = [474, 475, 476, 477]
+RIGHT_IRIS = [469, 470, 471, 472]
+
+overlay_dir = os.path.join(os.getcwd(),'overlays')
+overlay_files = list_overlay_images(overlay_dir)
+overlay_choices = [x.split('.png')[0] for x in overlay_files]
+with gr.Blocks() as demo:
+    with gr.Tab("Image"):
+        with gr.Row():
+            overlay_file = gr.Dropdown(choices=overlay_choices, value='Blue', label="Select a color")
+            # min_detection_confidence = gr.Slider(minimum=0.0, maximum=1.0, value=0.5, label="Min Detection Confidence")
+            # min_tracking_confidence = gr.Slider(minimum=0.0, maximum=1.0, value=0.5, label="Min Tracking Confidence")
+            # alpha = gr.Slider(minimum=0.0, maximum=1.0, value=0.3, label="Overlay Transparency")
+        with gr.Row():
+            input_image = gr.Image(height=500,width=400,label="Upload Image")
+            output_image = gr.Image(height=500,width=400,label="Processed Image")
+        process_image_btn = gr.Button("Process Image")
+        process_image_btn.click(process_image, 
+                                inputs=[input_image, overlay_file,], 
+                                outputs=output_image)
+    
+    with gr.Tab("Video"):
+        with gr.Row():
+            overlay_file = gr.Dropdown(choices=overlay_choices, value='Blue', label="Select a color")
+            # min_detection_confidence = gr.Slider(minimum=0.0, maximum=1.0, value=0.5, label="Min Detection Confidence")
+            # min_tracking_confidence = gr.Slider(minimum=0.0, maximum=1.0, value=0.5, label="Min Tracking Confidence")
+            # alpha = gr.Slider(minimum=0.0, maximum=1.0, value=0.3, label="Overlay Transparency")
+        with gr.Row():
+            input_video = gr.Video(height=500,width=400,label="Upload Video")
+            output_video = gr.Video(height=500,width=400,label="Processed Video")
+        process_video_btn = gr.Button("Process Video")
+        process_video_btn.click(process_video, 
+                                inputs=[input_video, overlay_file,], 
+                                outputs=output_video)
+
+    with gr.Tab("Webcam"):
+        with gr.Row():
+            overlay_file = gr.Dropdown(choices=overlay_choices, value='Blue', label="Select a color")
+            # min_detection_confidence = gr.Slider(minimum=0.0, maximum=1.0, value=0.5, label="Min Detection Confidence")
+            # min_tracking_confidence = gr.Slider(minimum=0.0, maximum=1.0, value=0.5, label="Min Tracking Confidence")
+            # alpha = gr.Slider(minimum=0.0, maximum=1.0, value=0.3, label="Overlay Transparency")
+        with gr.Row():
+            input_webcam = gr.Video(sources="webcam", label="Webcam")
+            output_webcam = gr.Image(label="Processed Webcam")
+        process_webcam_btn = gr.Button("Process Webcam")
+        process_webcam_btn.click(process_webcam, 
+                                inputs=[input_webcam, overlay_file,], 
+                                outputs=output_webcam)
+
+demo.launch()

requirements.txt

@@ -0,0 +1,4 @@
+gradio
+opencv-python
+opencv-python-headless
+mediapipe

utils.py

@@ -0,0 +1,114 @@
+import math
+import numpy as np
+import cv2 as cv
+
+def valid_float(n):
+    if not n.isfloat():
+        raise argparse.ArgumentTypeError('Invalid integer value: {}'.format(n))
+    return float(n)
+
+
+def euclaideanDistance(point, point1):
+    x, y = point
+    x1, y1 = point1
+    distance = math.sqrt((x1 - x)**2 + (y1 - y)**2)
+    return distance
+
+# Blinking Ratio
+def blinkRatio(img, landmarks, right_indices, left_indices):
+    # Right eyes 
+    # horizontal line 
+    rh_right = landmarks[right_indices[0]]
+    rh_left = landmarks[right_indices[8]]
+    # vertical line 
+    rv_top = landmarks[right_indices[12]]
+    rv_bottom = landmarks[right_indices[4]]
+    # draw lines on right eyes 
+    # cv.line(img, rh_right, rh_left, utils.GREEN, 2)
+    # cv.line(img, rv_top, rv_bottom, utils.WHITE, 2)    # LEFT_EYE 
+    # horizontal line 
+    lh_right = landmarks[left_indices[0]]
+    lh_left = landmarks[left_indices[8]]    # vertical line 
+    lv_top = landmarks[left_indices[12]]
+    lv_bottom = landmarks[left_indices[4]]    # Finding Distance Right Eye
+    rhDistance = euclaideanDistance(rh_right, rh_left)
+    rvDistance = euclaideanDistance(rv_top, rv_bottom)
+    # Finding Distance Left Eye
+    lvDistance = euclaideanDistance(lv_top, lv_bottom)
+    lhDistance = euclaideanDistance(lh_right, lh_left)    # Finding ratio of LEFT and Right Eyes
+    reRatio=0.0
+    leRatio=0.0
+    if (rvDistance > 0.0) & (lvDistance > 0.0):
+        reRatio = rhDistance/rvDistance
+        leRatio = lhDistance/lvDistance
+
+    ratio = (reRatio+leRatio)/2    
+    return ratio, reRatio, leRatio
+
+def process_frame(frame, overlay, LEFT_EYE, RIGHT_EYE, LEFT_IRIS, RIGHT_IRIS, 
+                  mp_face_mesh, min_detection_confidence, min_tracking_confidence,alpha):
+    with mp_face_mesh.FaceMesh(
+    max_num_faces=1,
+    refine_landmarks=True,
+    min_detection_confidence=min_detection_confidence,
+    min_tracking_confidence=min_tracking_confidence
+) as face_mesh:
+        # Convert frame to RGB
+        rgb_frame = cv.cvtColor(frame, cv.COLOR_BGR2RGB)
+        # Convert RGB frame to RGBA
+        rgba_frame = cv.cvtColor(frame, cv.COLOR_BGR2RGBA)
+        # Get frame dimensions
+        height, width = rgba_frame.shape[:2]
+        # Process frame with face mesh
+        results = face_mesh.process(rgb_frame)
+        if results.multi_face_landmarks:
+            # Initialize overlay with zeros
+            zero_overlay = np.zeros_like(rgba_frame)
+            # Get mesh points
+            mesh_points = np.array([np.multiply([p.x, p.y],
+                    [width, height]).astype(int) for p in results.multi_face_landmarks[0].landmark])
+            # Initialize iris masks
+            iris_mask_left = np.zeros(rgba_frame.shape, dtype=np.uint8)
+            iris_mask_right = np.zeros(rgba_frame.shape, dtype=np.uint8)
+            # Get blink ratio
+            _, re_ratio, le_ratio = blinkRatio(rgb_frame, mesh_points, RIGHT_EYE, LEFT_EYE)
+            # Get iris centers and radii
+            (l_cx, l_cy), l_radius = cv.minEnclosingCircle(mesh_points[LEFT_IRIS])
+            (r_cx, r_cy), r_radius = cv.minEnclosingCircle(mesh_points[RIGHT_IRIS])
+            center_left = (int(l_cx), int(l_cy))
+            center_right = (int(r_cx), int(r_cy))
+            # Draw circles on iris masks
+            cv.circle(iris_mask_left, center_left, int(l_radius), (255, 0, 0, 255), -1, cv.LINE_AA)
+            cv.circle(iris_mask_right, center_right, int(r_radius), (255, 0, 0, 255), -1, cv.LINE_AA)
+            # Get bounding box sizes
+            bbx_size_l = int((l_radius * 2) / 2)
+            bbx_size_r = int((r_radius * 2) / 2)
+            # Resize overlay
+            resized_overlay_l = cv.resize(overlay, (bbx_size_l * 2, bbx_size_l * 2), interpolation=cv.INTER_CUBIC)
+            resized_overlay_r = cv.resize(overlay, (bbx_size_r * 2, bbx_size_r * 2), interpolation=cv.INTER_CUBIC)
+            # Get bounding box coordinates
+            y1_r = center_right[1] - bbx_size_r
+            y2_r = center_right[1] + bbx_size_r
+            x1_r = center_right[0] - bbx_size_r
+            x2_r = center_right[0] + bbx_size_r
+            y1_l = center_left[1] - bbx_size_l
+            y2_l = center_left[1] + bbx_size_l
+            x1_l = center_left[0] - bbx_size_l
+            x2_l = center_left[0] + bbx_size_l
+            # Add resized overlay to zero overlay if conditions are met
+            if (resized_overlay_l.shape == zero_overlay[y1_l:y2_l, x1_l:x2_l].shape) & (le_ratio < 5.0) & (le_ratio > 2.0):
+                zero_overlay[y1_l:y2_l, x1_l:x2_l] = resized_overlay_l
+            if (resized_overlay_r.shape == zero_overlay[y1_r:y2_r, x1_r:x2_r].shape) & (re_ratio < 5.0) & (re_ratio > 2.0):
+                zero_overlay[y1_r:y2_r, x1_r:x2_r] = resized_overlay_r
+            # Initialize eye masks
+            eye_mask_left = np.zeros(rgba_frame.shape, dtype=np.uint8)
+            eye_mask_right = np.zeros(rgba_frame.shape, dtype=np.uint8)
+            # Fill eye masks with polygons
+            cv.fillPoly(eye_mask_left, [mesh_points[LEFT_EYE]], (255, 0, 0, 255))
+            cv.fillPoly(eye_mask_right, [mesh_points[RIGHT_EYE]], (255, 0, 0, 255))
+            # Use the 4-channel masks to create zero_overlay
+            zero_overlay[np.where((iris_mask_left[:, :, 3] > 0) & (eye_mask_left[:, :, 3] == 0))] = 0
+            zero_overlay[np.where((iris_mask_right[:, :, 3] > 0) & (eye_mask_right[:, :, 3] == 0))] = 0
+            # Add weighted overlay to frame
+            rgba_frame = cv.addWeighted(rgba_frame, 1, zero_overlay, alpha, 0)
+        return rgba_frame