Create app.py

app.py

@@ -0,0 +1,54 @@
+import gradio as gr
+import torch
+import torchvision
+from torchvision import transforms
+import numpy as np
+from PIL import Image
+
+# Use GPU if available
+device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+torch.backends.cudnn.benchmark = True
+print(f"Using device: {device}")
+
+# Load lightweight detection model
+model = torchvision.models.detection.fasterrcnn_mobilenet_v3_large_fpn(pretrained=True)
+model.to(device)
+model.eval()
+
+# Image transformation
+transform = transforms.Compose([
+    transforms.ToTensor(),
+])
+
+# Mixed precision (FP16) for CUDA
+autocast = torch.cuda.amp.autocast if device.type == "cuda" else torch.cpu.amp.autocast
+
+def count_persons(image):
+    # Convert image to tensor
+    image_rgb = np.array(image.convert("RGB"))
+    img_tensor = transform(image_rgb).to(device).unsqueeze(0)
+
+    # Inference
+    with torch.no_grad():
+        with autocast():
+            outputs = model(img_tensor)[0]
+
+    # Count persons (label 1 in COCO)
+    person_count = sum(
+        1 for label, score in zip(outputs['labels'], outputs['scores'])
+        if label.item() == 1 and score.item() > 0.8
+    )
+
+    return f"Number of persons detected: {person_count}"
+
+# Gradio interface for image upload
+demo = gr.Interface(
+    fn=count_persons,
+    inputs=gr.Image(type="pil", label="Upload Image"),
+    outputs=gr.Text(label="Person Count"),
+    title="Person Counter in Image (Fast)",
+    description="Upload an image to count the number of people using a fast MobileNet-based detector. GPU supported."
+)
+
+if __name__ == "__main__":
+    demo.launch()