Spaces:
Sleeping
Sleeping
File size: 3,169 Bytes
3377ea4 446c2df 3377ea4 1a8adab 3377ea4 1a8adab 3377ea4 446c2df |
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 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 |
import gradio as gr
import torch
import torchvision.transforms as transforms
from PIL import Image
import os
from pathlib import Path
class FoodImageClassifier:
def __init__(self, model_dir="traced_models/food_101_vit_small",
model_file_name="model.pt",
labels_path='food_101_classes.txt'):
self.device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
print(self.device)
# Load the traced model
model_full_path = Path(model_dir,model_file_name)
self.model = torch.jit.load(model_full_path)
self.model = self.model.to(self.device)
self.model.eval()
# Define the same transforms used during training/testing
self.transforms = transforms.Compose([
transforms.Resize(256),
transforms.CenterCrop(224),
transforms.ToTensor(),
transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])
])
# Load labels from file
with open(labels_path, 'r') as f:
self.labels = [line.strip() for line in f.readlines()]
@torch.no_grad()
def predict(self, image):
if image is None:
return None
# Convert to PIL Image if needed
if not isinstance(image, Image.Image):
image = Image.fromarray(image).convert('RGB')
# Preprocess image
img_tensor = self.transforms(image).unsqueeze(0).to(self.device)
# Get prediction
output = self.model(img_tensor)
probabilities = torch.nn.functional.softmax(output[0], dim=0)
# Create prediction dictionary
return {
self.labels[idx]: float(prob)
for idx, prob in enumerate(probabilities)
}
# Create classifier instance
classifier = FoodImageClassifier()
# Format available classes into HTML table - 10 per row
formatted_classes = ['<tr>']
for i, label in enumerate(classifier.labels):
if i > 0 and i % 10 == 0:
formatted_classes.append('</tr><tr>')
formatted_classes.append(f'<td>{label}</td>')
formatted_classes.append('</tr>')
# Create HTML table with styling
table_html = f"""
<style>
.food-classes-table {{
width: 100%;
border-collapse: collapse;
margin: 10px 0;
}}
.food-classes-table td {{
padding: 6px;
text-align: center;
border: 1px solid var(--border-color-primary);
font-size: 14px;
color: var(--body-text-color);
}}
.food-classes-table tr td {{
background-color: var(--background-fill-primary);
}}
</style>
<table class="food-classes-table">
{''.join(formatted_classes)}
</table>
"""
# Create Gradio interface
demo = gr.Interface(
fn=classifier.predict,
inputs=gr.Image(),
outputs=gr.Label(num_top_classes=5),
title="Food classifier",
description="Upload an image to classify Food Images",
examples=[
["sample_data/apple_pie.jpg"],
["sample_data/pizza.jpg"]
],
article=f"Available food classes:\n{table_html}"
)
if __name__ == "__main__":
demo.launch(share=True) |