Update app.py

app.py

@@ -1,81 +1,63 @@
-import gradio as gr
 import cv2
 import mediapipe as mp
+import streamlit as st
 import tempfile
+import os
 
-# Initialize MediaPipe once
-mp_hands = mp.solutions.hands
-mp_pose = mp.solutions.pose
+# mediapipe pose setup
 mp_drawing = mp.solutions.drawing_utils
+mp_pose = mp.solutions.pose
 
-hands = mp_hands.Hands(
-    static_image_mode=False,
-    max_num_hands=2,
-    min_detection_confidence=0.5,
-    min_tracking_confidence=0.5
-)
-
-pose = mp_pose.Pose(
-    static_image_mode=False,
-    model_complexity=1,
-    enable_segmentation=False,
-    min_detection_confidence=0.5,
-    min_tracking_confidence=0.5
-)
+st.title("MediaPipe Body Pose Detection")
 
-def process_video(video_path, target_width=640):
-    cap = cv2.VideoCapture(video_path)
-    fourcc = cv2.VideoWriter_fourcc(*'mp4v')
-    temp_output = tempfile.NamedTemporaryFile(delete=False, suffix=".mp4")
-    out = None
+# select input source
+option = st.radio("choose input source", ("webcam", "upload video"))
 
-    while cap.isOpened():
-        ret, frame = cap.read()
-        if not ret:
-            break
+video_path = None
+if option == "upload video":
+    uploaded_file = st.file_uploader("upload a video file", type=["mp4", "mov", "avi"])
+    if uploaded_file:
+        temp_file = tempfile.NamedTemporaryFile(delete=False)
+        temp_file.write(uploaded_file.read())
+        video_path = temp_file.name
 
-        h0, w0 = frame.shape[:2]
-        # Resize frame keeping aspect ratio
-        scale = target_width / float(w0)
-        target_height = int(round(h0 * scale))
-        frame_small = cv2.resize(frame, (target_width, target_height), interpolation=cv2.INTER_AREA)
+elif option == "webcam":
+    video_path = 0  # 0 for default webcam
 
-        # Convert BGR to RGB
-        rgb_small = cv2.cvtColor(frame_small, cv2.COLOR_BGR2RGB)
+if video_path is not None:
+    stframe = st.empty()
 
-        # Hand detection
-        hand_results = hands.process(rgb_small)
-        if hand_results.multi_hand_landmarks:
-            for hand_landmarks in hand_results.multi_hand_landmarks:
-                mp_drawing.draw_landmarks(frame_small, hand_landmarks, mp_hands.HAND_CONNECTIONS)
+    cap = cv2.VideoCapture(video_path)
 
-        # Pose detection
-        pose_results = pose.process(rgb_small)
-        if pose_results.pose_landmarks:
-            mp_drawing.draw_landmarks(frame_small, pose_results.pose_landmarks, mp_pose.POSE_CONNECTIONS)
+    with mp_pose.Pose(min_detection_confidence=0.5, min_tracking_confidence=0.5) as pose:
+        while cap.isOpened():
+            ret, frame = cap.read()
+            if not ret:
+                break
 
-        # Initialize video writer
-        if out is None:
-            out = cv2.VideoWriter(temp_output.name, fourcc, cap.get(cv2.CAP_PROP_FPS), (frame_small.shape[1], frame_small.shape[0]))
+            # convert to rgb for processing
+            image = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
+            image.flags.writeable = False
 
-        out.write(frame_small)
+            results = pose.process(image)
 
-    cap.release()
-    if out:
-        out.release()
+            # convert back to BGR for display
+            image.flags.writeable = True
+            image = cv2.cvtColor(image, cv2.COLOR_RGB2BGR)
 
-    return temp_output.name
+            # draw landmarks and connections if detected
+            if results.pose_landmarks:
+                mp_drawing.draw_landmarks(
+                    image,
+                    results.pose_landmarks,
+                    mp_pose.POSE_CONNECTIONS,
+                    mp_drawing.DrawingSpec(color=(0,0,255), thickness=2, circle_radius=2),  # landmarks
+                    mp_drawing.DrawingSpec(color=(0,255,0), thickness=2, circle_radius=2)   # connections
+                )
 
-iface = gr.Interface(
-    fn=process_video,
-    inputs=[
-        gr.Video(label="Upload a video"),
-        gr.Slider(minimum=160, maximum=1280, value=640, step=20, label="Processing width")
-    ],
-    outputs=gr.Video(label="Processed video"),
-    title="Hand & Body Pose Detection from Video",
-    description="Upload a video and see MediaPipe detect hands and body pose."
-)
+            # display frame in streamlit
+            stframe.image(image, channels="BGR")
 
-if __name__ == "__main__":
-    iface.launch()
+    cap.release()
+    if option == "upload video":
+        os.remove(video_path)