Upload 5 files

app.py.py

@@ -0,0 +1,60 @@
+import gradio as gr
+import cv2
+import os
+from ultralytics import YOLO
+from helmet_detect_alert import detect_from_images, detect_and_alert, MODEL_PATHS, CONFIDENCE_THRESHOLD
+
+model_cache = {}
+
+def load_model(model_name):
+    if model_name not in model_cache:
+        model_cache[model_name] = YOLO(MODEL_PATHS[model_name])
+    return model_cache[model_name]
+
+def process_image(image, model_name, confidence):
+    model = load_model(model_name)
+    results = model.predict(source=image, conf=confidence, verbose=False)
+    annotated = results[0].plot()
+    return annotated
+
+def process_video(video, model_name, confidence):
+    input_path = "temp_input_video.mp4"
+    output_path = "temp_output_video.mp4"
+    with open(input_path, "wb") as f:
+        f.write(video.read())
+    model = load_model(model_name)
+    detect_and_alert(input_path, output_path, model, confidence)
+    return output_path
+
+with gr.Blocks() as demo:
+    gr.Markdown("# 🪖 Helmet Detection App (YOLOv8)")
+
+    with gr.Tab("📷 Image Detection"):
+        with gr.Row():
+            image_input = gr.Image(type="numpy", label="Upload Image")
+            image_output = gr.Image(label="Detection Result")
+
+        with gr.Row():
+            conf_slider_img = gr.Slider(0.1, 1.0, value=CONFIDENCE_THRESHOLD, label="Confidence Threshold")
+            model_selector_img = gr.Dropdown(choices=list(MODEL_PATHS.keys()), value="YOLOv8n", label="YOLO Model")
+            btn_detect_img = gr.Button("🔍 Detect Helmet")
+
+        btn_detect_img.click(fn=process_image, inputs=[image_input, model_selector_img, conf_slider_img], outputs=image_output)
+
+    with gr.Tab("🎥 Video Detection"):
+        with gr.Row():
+            video_input = gr.File(file_types=[".mp4"], label="Upload Video")
+            video_output = gr.Video(label="Detection Result")
+
+        with gr.Row():
+            conf_slider_vid = gr.Slider(0.1, 1.0, value=CONFIDENCE_THRESHOLD, label="Confidence Threshold")
+            model_selector_vid = gr.Dropdown(choices=list(MODEL_PATHS.keys()), value="YOLOv8n", label="YOLO Model")
+            btn_detect_vid = gr.Button("🎬 Detect Helmet in Video")
+
+        btn_detect_vid.click(fn=process_video, inputs=[video_input, model_selector_vid, conf_slider_vid], outputs=video_output)
+
+    gr.Markdown("---")
+    gr.Markdown("Developed by **Kiran Subedi** | kiransubedi545@gmail.com | [Website](http://kiransubedi545.com.np)")
+
+if __name__ == "__main__":
+    demo.launch()

best.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4b87f454e987b44a6d2baa1286aaae7562a6db1e6f81a407b8cc9ed190d7a4be
+size 6213219

helmet_detect_alert.py

@@ -0,0 +1,140 @@
+import cv2
+import os
+import pyttsx3
+from ultralytics import YOLO
+import numpy as np
+from tkinter import filedialog, messagebox
+import tkinter as tk
+from PIL import Image, ImageTk
+import csv
+from gtts import gTTS
+import tempfile
+import pygame
+from datetime import datetime
+
+# ----------------- Configuration -----------------
+DOWNLOAD_PATH = r"D:\Learning Programming\Python\Deep learning\YOLOv8\Test_videos"
+MODEL_PATHS = {
+    "YOLOv8n": r"D:\Learning Programming\Python\Deep learning\YOLOv8\runs\detect\train5\weights\best.pt"
+}
+ALERT_ICON_PATH = "alert_icon.png"
+ALERT_LOG_PATH = "alert_log.csv"
+CONFIDENCE_THRESHOLD = 0.3
+ALERT_LANGUAGE = "en"
+
+# ----------------- Initialize Modules -----------------
+pygame.init()
+os.makedirs(DOWNLOAD_PATH, exist_ok=True)
+
+# ----------------- Add Alert Symbol -----------------
+def overlay_alert_icon(frame, x, y):
+    if not os.path.exists(ALERT_ICON_PATH):
+        return frame
+    icon = cv2.imread(ALERT_ICON_PATH, cv2.IMREAD_UNCHANGED)
+    if icon is None:
+        return frame
+    icon = cv2.resize(icon, (50, 50))
+    ih, iw = icon.shape[:2]
+    if y+ih > frame.shape[0] or x+iw > frame.shape[1]:
+        return frame
+    alpha_s = icon[:, :, 3] / 255.0
+    alpha_l = 1.0 - alpha_s
+    for c in range(3):
+        frame[y:y+ih, x:x+iw, c] = (alpha_s * icon[:, :, c] + alpha_l * frame[y:y+ih, x:x+iw, c])
+    return frame
+
+# ----------------- Voice Alert -----------------
+def speak_alert(text, lang="en"):
+    with tempfile.NamedTemporaryFile(delete=True) as fp:
+        tts = gTTS(text=text, lang=lang)
+        temp_path = f"{fp.name}.mp3"
+        tts.save(temp_path)
+        pygame.mixer.music.load(temp_path)
+        pygame.mixer.music.play()
+        while pygame.mixer.music.get_busy():
+            continue
+
+# ----------------- Save Alert Log -----------------
+def log_alert(message):
+    timestamp = datetime.now().strftime("%Y-%m-%d %H:%M:%S")
+    with open(ALERT_LOG_PATH, mode='a', newline='') as file:
+        writer = csv.writer(file)
+        writer.writerow([timestamp, message])
+
+# ----------------- Run YOLO Prediction + Alerts -----------------
+def detect_and_alert(video_path, model, confidence=CONFIDENCE_THRESHOLD):
+    cap = cv2.VideoCapture(video_path)
+    frame_count = 0
+    alert_cooldown = 15
+    while cap.isOpened():
+        ret, frame = cap.read()
+        if not ret:
+            break
+        frame_count += 1
+        if frame_count % 3 != 0:
+            continue
+        results = model.predict(source=frame, conf=confidence, verbose=False)
+        annotated_frame = results[0].plot()
+        helmets, heads = [], []
+        for box in results[0].boxes:
+            cls_id = int(box.cls[0].item())
+            x1, y1, x2, y2 = box.xyxy[0].cpu().numpy().astype(int)
+            center_x = (x1 + x2) // 2
+            center_y = (y1 + y2) // 2
+            if cls_id == 0:
+                helmets.append(((x1, y1, x2, y2), (center_x, center_y)))
+            elif cls_id == 2:
+                heads.append(((x1, y1, x2, y2), (center_x, center_y)))
+        alert_triggered = False
+        for (hx1, hy1, hx2, hy2), (hcx, hcy) in heads:
+            covered = any(abs(hcx - hel_x) < 30 and abs(hcy - hel_y) < 30 for (_, _, _, _), (hel_x, hel_y) in helmets)
+            if not covered:
+                annotated_frame = overlay_alert_icon(annotated_frame, hx1, hy1)
+                cv2.putText(annotated_frame, "⚠ No Helmet!", (hx1, hy1 - 10), cv2.FONT_HERSHEY_SIMPLEX, 0.7, (0, 0, 255), 2)
+                alert_triggered = True
+                log_alert(f"Alert at frame {frame_count}: No helmet at ({hx1}, {hy1})")
+        if alert_triggered and frame_count % alert_cooldown == 0:
+            speak_alert("Alert! Person without helmet detected", ALERT_LANGUAGE)
+        cv2.imshow("Helmet Detection", annotated_frame)
+        if cv2.waitKey(1) & 0xFF == ord('q'):
+            break
+    cap.release()
+    cv2.destroyAllWindows()
+
+# ----------------- Detect from Multiple Images -----------------
+def detect_from_images(image_paths, model, confidence=CONFIDENCE_THRESHOLD):
+    for image_path in image_paths:
+        frame = cv2.imread(image_path)
+        results = model.predict(source=frame, conf=confidence, verbose=False)
+        annotated_frame = results[0].plot()
+        helmets, heads = [], []
+        for box in results[0].boxes:
+            cls_id = int(box.cls[0].item())
+            x1, y1, x2, y2 = box.xyxy[0].cpu().numpy().astype(int)
+            center_x = (x1 + x2) // 2
+            center_y = (y1 + y2) // 2
+            if cls_id == 0:
+                helmets.append(((x1, y1, x2, y2), (center_x, center_y)))
+            elif cls_id == 2:
+                heads.append(((x1, y1, x2, y2), (center_x, center_y)))
+        alert_triggered = False
+        for (hx1, hy1, hx2, hy2), (hcx, hcy) in heads:
+            covered = any(abs(hcx - hel_x) < 30 and abs(hcy - hel_y) < 30 for (_, _, _, _), (hel_x, hel_y) in helmets)
+            if not covered:
+                annotated_frame = overlay_alert_icon(annotated_frame, hx1, hy1)
+                cv2.putText(annotated_frame, "⚠ No Helmet!", (hx1, hy1 - 10), cv2.FONT_HERSHEY_SIMPLEX, 0.7, (0, 0, 255), 2)
+                alert_triggered = True
+                log_alert(f"Alert: No helmet at ({hx1}, {hy1}) in image {image_path}")
+        if alert_triggered:
+            speak_alert("Alert! Person without helmet detected", ALERT_LANGUAGE)
+        cv2.imshow(f"Result - {os.path.basename(image_path)}", annotated_frame)
+        cv2.waitKey(0)
+    cv2.destroyAllWindows()
+
+# ----------------- Snapshot -----------------
+def save_snapshot(frame, path="snapshot.jpg"):
+    cv2.imwrite(path, frame)
+
+# ----------------- Main Execution -----------------
+model = YOLO(MODEL_PATHS["YOLOv8n"])
+# This script is now module-only: import and use functions from app.py

requirements.txt

@@ -0,0 +1,5 @@
+ultralytics
+opencv-python
+Pillow
+gTTS
+pygame