Upload 4 files

app_gradio.py

@@ -0,0 +1,91 @@
+import pickle
+import cv2
+import mediapipe as mp
+import numpy as np
+import matplotlib.pyplot as plt
+import gradio as gr
+import time
+from PIL import Image
+import random
+
+# Load model once globally
+model_dict = pickle.load(open('./model.p', 'rb'))
+model = model_dict['model']
+
+mp_hands = mp.solutions.hands
+mp_drawing = mp.solutions.drawing_utils
+mp_drawing_styles = mp.solutions.drawing_styles
+
+# Function to process a single frame
+def predict(input):
+    hands = mp_hands.Hands(static_image_mode=True, min_detection_confidence=0.3)
+    labels_dict =  {0: 'L', 1: 'A', 2: 'B', 3: 'C', 4: 'V', 5: 'W', 6: 'Y'}
+    data_aux = []
+    frame_rgb = cv2.cvtColor(input, cv2.COLOR_BGR2RGB)
+    results = hands.process(frame_rgb)
+    
+    if results.multi_hand_landmarks:
+        for hand_landmarks in results.multi_hand_landmarks:
+            mp_drawing.draw_landmarks(
+                input,
+                hand_landmarks,
+                mp_hands.HAND_CONNECTIONS,
+                mp_drawing_styles.get_default_hand_landmarks_style(),
+                mp_drawing_styles.get_default_hand_connections_style())
+
+            for i in range(len(hand_landmarks.landmark)):
+                x = hand_landmarks.landmark[i].x
+                y = hand_landmarks.landmark[i].y
+                data_aux.append(x)
+                data_aux.append(y)
+    
+    if data_aux:
+        prediction = model.predict([np.asarray(data_aux)])
+        predicted_character = labels_dict[int(prediction[0])]
+        output = cv2.putText(input, f'Sign: {predicted_character}', (5, 30), cv2.FONT_HERSHEY_SIMPLEX, 2, (255, 255, 255), 3, cv2.LINE_AA)
+        output_image = Image.fromarray(cv2.cvtColor(output, cv2.COLOR_BGR2RGB))
+        return output_image, predicted_character
+    else:
+        return Image.fromarray(cv2.cvtColor(input, cv2.COLOR_BGR2RGB)), "No hand landmarks detected in the current frame."
+
+# Function to process video
+def vid_inf(vid, progress=gr.Progress()):
+    cap = cv2.VideoCapture(vid)
+    total_frames = int(cap.get(cv2.CAP_PROP_FRAME_COUNT))
+    processed_frames = 0
+    writer = None
+    character_output = []
+    tmpname = f"output_{random.randint(111111111, 999999999)}.mp4"
+    
+    while cap.isOpened():
+        ret, frame = cap.read()
+        if not ret:
+            break
+        
+        if writer is None:
+            height, width, _ = frame.shape
+            writer = cv2.VideoWriter(tmpname, cv2.VideoWriter_fourcc(*'mp4v'), cap.get(cv2.CAP_PROP_FPS), (width, height))
+        
+        processed_frames += 1
+        progress(processed_frames / total_frames, desc=f"Processing frame {processed_frames}/{total_frames}")
+        out, character = predict(frame)
+        writer.write(cv2.cvtColor(np.array(out), cv2.COLOR_RGB2BGR))
+        character_output.append(character)
+    
+    cap.release()
+    writer.release()
+    return tmpname, character_output
+
+input_video = gr.Video(sources=["webcam", "upload"], label="Input Video")
+output_video = gr.Video(label="Processed Video")
+output_character = gr.Textbox(label="Sign Sequence")
+
+# Create Gradio Interface for Video Inference
+interface_video = gr.Interface(
+    fn=vid_inf,
+    inputs=[input_video],
+    outputs=[output_video, output_character],
+    title="Video Inference"
+)
+
+interface_video.launch(share=True)

packages.txt

@@ -0,0 +1 @@
+ffmpeg

requirements.txt

@@ -0,0 +1,7 @@
+scikit-learn==1.2.0
+mediapipe==0.10.9
+numpy==1.24.3
+matplotlib==3.7.2
+gradio==3.1.4
+opencv-python-headless==4.5.5.62
+Pillow==8.3.2