File size: 1,788 Bytes
6265901 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 |
import gradio as gr
from PIL import Image
import torch
import torchvision.transforms as transforms
from model import RetinaNet # Import your RetinaNet model definition
# Define the image transformation pipeline
image_transform = transforms.Compose([
transforms.Resize((224, 224)),
transforms.ToTensor(),
transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])
])
# Load the model
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
model = RetinaNet(num_classes=2).to(device)
model.load_state_dict(torch.load("retinanet_best_model.pth", map_location=device))
model.eval()
# Prediction function
def predict_image(image):
# Preprocess the image
img = Image.fromarray(image).convert('RGB') # Convert Gradio input to PIL Image
input_tensor = image_transform(img).unsqueeze(0).to(device)
# Perform inference
with torch.no_grad():
prediction = model(input_tensor.float())
sum_value = abs(torch.sum(prediction[0]))
p_true = abs(prediction[0][0])
p_false = abs(prediction[0][1])
# Interpret the prediction
if p_true > 0.7:
result = "Accepted"
confidence = float(p_true)
else:
result = "Rejected"
confidence = float(p_false)
return f"Result: {result}, Confidence: {confidence:.2f}"
# Create the Gradio interface
with gr.Blocks() as demo:
gr.Markdown("# RetinaNet Model Prediction")
with gr.Row():
image_input = gr.Image(label="Upload Image", type="numpy")
output_text = gr.Textbox(label="Prediction Result")
predict_button = gr.Button("Predict")
predict_button.click(predict_image, inputs=image_input, outputs=output_text)
# Launch the app
demo.launch() |