Update app.py

app.py

@@ -6,57 +6,66 @@ import tensorflow as tf
 import streamlit as st 
 import tempfile
 
+# Function to detect hand using Haar Cascade
+def detect_hand(frame, hand_cascade):
+    gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
+    hands = hand_cascade.detectMultiScale(gray, scaleFactor=1.1, minNeighbors=5, minSize=(30, 30))
+    return hands
+
+# Load Haar Cascade for hand detection
+hand_cascade_path = 'path/to/your/hand_cascade.xml'  # Replace with your actual path
+hand_cascade = cv2.CascadeClassifier(hand_cascade_path)
+
 # Open the video file
 f = st.file_uploader("Choose a Video")
 
 if f is not None:
-    # Read the video file from the file-like object
-          
     tfile = tempfile.NamedTemporaryFile(delete=False) 
     tfile.write(f.read())
-
-# Opens the Video file
-    cap= cv2.VideoCapture(tfile.name)
-
-    # Get the frames per second (fps) of the video
-    fps = (cap.get(cv2.CAP_PROP_FPS)) 
+    cap = cv2.VideoCapture(tfile.name)
+    fps = cap.get(cv2.CAP_PROP_FPS)
     st.write(fps)
-
-    # Calculate the interval to capture one frame per second
     interval = int(round(fps/1))
-
-    # Initialize a counter for frames
     frame_count = 0
     model = tf.keras.models.load_model('HandSignClassifier (1).h5')
     array = ['a','b','c','d','e','f','g','h','i','k','l','m','n','o','p','q','r','s','t','u','v','w','x','y']
     out = ''
-    
+
     while True:
-        # Read the next fram
-        
         ret, frame = cap.read()
-
-        # Break the loop if the video is over
         if not ret:
             break
 
         # Check if it's time to capture a frame
         if frame_count % interval == 0:
-            frame = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)  # Convert to grayscale
-            frame = cv2.resize(frame, (28, 28))  # Resize to (28, 28)
-            frame = np.reshape(frame, (1, 28, 28, 1))
-            st.image(frame, 'input')# Reshape
-            pred = model.predict(frame) 
-            st.write(pred)
-            pred = np.argmax(pred)
-            pred = array[pred]
-            if not out or out[-1] != pred:
-                out = out + pred
+            hands = detect_hand(frame, hand_cascade)
+
+            for (x, y, w, h) in hands:
+                # Draw rectangles around detected hands
+                cv2.rectangle(frame, (x, y), (x+w, y+h), (255, 0, 0), 2)
+
+            # Display the frame with detected hands
+            st.image(frame, 'input')
+
+            if hands:
+                # Extract the region of interest for hand from the frame
+                hand_roi = frame[y:y+h, x:x+w]
+
+                # Preprocess the hand ROI for your model (resize, convert to grayscale, etc.)
+                hand_roi = cv2.cvtColor(hand_roi, cv2.COLOR_BGR2GRAY)
+                hand_roi = cv2.resize(hand_roi, (28, 28))
+                hand_roi = np.reshape(hand_roi, (1, 28, 28, 1))
+
+                # Make predictions using your model
+                pred = model.predict(hand_roi)
+                st.write(pred)
+                pred = np.argmax(pred)
+                pred = array[pred]
+                if not out or out[-1] != pred:
+                    out = out + pred
 
         # Increment the frame counter
         frame_count += 1
 
-    # Release the video capture object
     cap.release()
-
     st.write(out)