File size: 5,973 Bytes
8e8eb27 |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 |
import gradio as gr
import time
import torch
from transformers import pipeline
from PIL import Image
import pandas as pd
import matplotlib.pyplot as plt
import io
# --- 1. Model Configuration & Metadata ---
DEVICE = "cuda:0" if torch.cuda.is_available() else "cpu"
MODEL_INFO = {
"ViT (eslamxm/vit-base-food101)": {
"model_id": "eslamxm/vit-base-food101",
"benchmark_accuracy": 90.68,
"pipeline": None
},
"Swin (aspis/swin-finetuned-food101)": {
"model_id": "aspis/swin-finetuned-food101",
"benchmark_accuracy": 93.81,
"pipeline": None
}
}
# --- 2. Lazy Loading of Models ---
def load_pipeline(model_name):
"""Loads a model pipeline only when it's first needed."""
if MODEL_INFO[model_name]["pipeline"] is None:
print(f"Loading model: {model_name}...")
model_id = MODEL_INFO[model_name]["model_id"]
MODEL_INFO[model_name]["pipeline"] = pipeline(task="image-classification", model=model_id, device=DEVICE)
print(f"Model '{model_name}' loaded on {DEVICE}.")
return MODEL_INFO[model_name]["pipeline"]
# --- 3. Function to Generate Comparison Chart ---
def create_comparison_chart(selected_model_name, current_inference_time):
"""Generates a bar chart comparing model accuracy and inference time."""
data = {'Model': [], 'Metric': [], 'Value': []}
for name, info in MODEL_INFO.items():
data['Model'].append(name)
data['Metric'].append('Benchmark Accuracy (%)')
data['Value'].append(info['benchmark_accuracy'])
data['Model'].append(selected_model_name)
data['Metric'].append('Current Inference Time (s)')
data['Value'].append(current_inference_time)
df = pd.DataFrame(data)
fig, ax = plt.subplots(1, 2, figsize=(12, 5))
fig.suptitle('Model Performance Comparison', fontsize=16)
acc_df = df[df['Metric'] == 'Benchmark Accuracy (%)']
colors_acc = ['#4c72b0' if model != selected_model_name else '#2ca02c' for model in acc_df['Model']]
acc_plot = acc_df.plot(kind='bar', x='Model', y='Value', ax=ax[0], color=colors_acc, legend=None)
ax[0].set_title('Benchmark Accuracy')
ax[0].set_ylabel('Accuracy (%)')
ax[0].set_xlabel('')
ax[0].set_ylim(0, 100)
ax[0].tick_params(axis='x', rotation=10)
for p in acc_plot.patches:
ax[0].annotate(f"{p.get_height():.2f}%", (p.get_x() + p.get_width() / 2., p.get_height()),
ha='center', va='center', xytext=(0, 9), textcoords='offset points')
time_df = df[df['Metric'] == 'Current Inference Time (s)']
time_plot = time_df.plot(kind='bar', x='Model', y='Value', ax=ax[1], color=['#d62728'])
ax[1].set_title('Inference Time for This Image')
ax[1].set_ylabel('Time (seconds)')
ax[1].set_xlabel('')
ax[1].tick_params(axis='x', rotation=0)
for p in time_plot.patches:
ax[1].annotate(f"{p.get_height():.4f}s", (p.get_x() + p.get_width() / 2., p.get_height()),
ha='center', va='center', xytext=(0, 9), textcoords='offset points')
plt.tight_layout(rect=[0, 0.03, 1, 0.95])
return fig
# --- 4. The Core Classification Function ---
def classify_image(image, model_name):
"""
Takes an image and model name, returns predictions, inference time,
and a comparison chart.
"""
if image is None:
return {}, "Please upload an image first.", None, "Please upload an image to see a comparison."
pipe = load_pipeline(model_name)
start_time = time.time()
predictions = pipe(Image.fromarray(image))
end_time = time.time()
inference_time = end_time - start_time
top_5_preds = {p['label'].replace("_", " ").title(): p['score'] for p in predictions[:5]}
comparison_fig = create_comparison_chart(model_name, inference_time)
buf = io.BytesIO()
comparison_fig.savefig(buf, format='png', bbox_inches='tight')
buf.seek(0)
comparison_img = Image.open(buf)
plt.close(comparison_fig)
return (
top_5_preds,
f"Inference Time: {inference_time:.4f} seconds",
comparison_img,
f"Chart shows accuracy for all models and the inference time for the **{model_name}** model on this specific image."
)
# --- 5. Gradio Interface Definition ---
with gr.Blocks(theme=gr.themes.Soft(), css="footer {display: none !important}") as demo:
gr.Markdown("# 🍔 Food Classifier: Accuracy vs. Speed")
gr.Markdown(
"Compare two different models for classifying food images from the Food101 dataset. "
"Notice the trade-off: the **Swin** model is more accurate but might be slower, while the **ViT** model is faster but slightly less accurate."
)
with gr.Row(variant="panel"):
with gr.Column(scale=1):
image_input = gr.Image(type="numpy", label="Upload a food picture")
model_dropdown = gr.Dropdown(
choices=list(MODEL_INFO.keys()),
value=list(MODEL_INFO.keys())[0],
label="Choose a Model"
)
classify_button = gr.Button("Classify Image", variant="primary")
gr.Examples(
examples=[
["examples/sushi.jpg", list(MODEL_INFO.keys())[1]],
["examples/pizza.jpg", list(MODEL_INFO.keys())[0]],
["examples/apple_pie.jpg", list(MODEL_INFO.keys())[1]],
],
inputs=[image_input, model_dropdown],
)
with gr.Column(scale=2):
output_label = gr.Label(num_top_classes=5, label="Top 5 Predictions")
output_time = gr.Textbox(label="Performance")
output_chart = gr.Image(type="pil", label="Model Comparison Chart")
chart_info = gr.Markdown()
classify_button.click(
fn=classify_image,
inputs=[image_input, model_dropdown],
outputs=[output_label, output_time, output_chart, chart_info]
)
if __name__ == "__main__":
demo.launch()
|