Image Classification
Transformers
ONNX
English
multi-head-classification
room-classification
dinov2
computer-vision
scene-classification
Instructions to use ondame/image-classifier with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use ondame/image-classifier with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("image-classification", model="ondame/image-classifier") pipe("https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/hub/parrots.png")# Load model directly from transformers import AutoModel model = AutoModel.from_pretrained("ondame/image-classifier", device_map="auto") - Notebooks
- Google Colab
- Kaggle
File size: 5,464 Bytes
0812af4 575aa0e 0812af4 575aa0e 0812af4 575aa0e 0812af4 575aa0e 0812af4 575aa0e 0812af4 575aa0e | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 | #!/usr/bin/env python3
"""
ONNX ๋ชจ๋ธ์ ์ฌ์ฉํ ๋ฉํฐํค๋ ์ด๋ฏธ์ง ๋ถ๋ฅ ์ถ๋ก ์์
์ ์ฒด ๋ชจ๋ธ(model.onnx) ๋๋ ๋ถ๋ฆฌ ๋ชจ๋ธ(encoder.onnx + head.onnx) ์ฌ์ฉ ๊ฐ๋ฅ
"""
import onnxruntime as ort
import numpy as np
from PIL import Image
import torchvision.transforms as transforms
import json
from pathlib import Path
# ์ ์ฒ๋ฆฌ ํ์ดํ๋ผ์ธ
transform = transforms.Compose([
transforms.Resize((224, 224)),
transforms.CenterCrop(224),
transforms.ToTensor(),
transforms.Normalize(mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225])
])
def load_model_info(model_info_path):
"""๋ชจ๋ธ ์ ๋ณด ๋ก๋"""
with open(model_info_path, 'r', encoding='utf-8') as f:
return json.load(f)
def preprocess_image(image_path):
"""์ด๋ฏธ์ง ์ ์ฒ๋ฆฌ"""
image = Image.open(image_path).convert('RGB')
tensor = transform(image)
return tensor.unsqueeze(0).numpy() # ๋ฐฐ์น ์ฐจ์ ์ถ๊ฐ
def softmax(x):
"""Softmax ํจ์"""
exp_x = np.exp(x - np.max(x, axis=1, keepdims=True))
return exp_x / np.sum(exp_x, axis=1, keepdims=True)
def predict_image_full_model(model_path, model_info_path, image_path):
"""์ ์ฒด ๋ชจ๋ธ์ ์ฌ์ฉํ ์ด๋ฏธ์ง ๋ถ๋ฅ ์์ธก"""
# ๋ชจ๋ธ ์ ๋ณด ๋ก๋
model_info = load_model_info(model_info_path)
# ONNX ์ธ์
์์ฑ
session = ort.InferenceSession(model_path)
# ์ด๋ฏธ์ง ์ ์ฒ๋ฆฌ
image_array = preprocess_image(image_path)
# ์ถ๋ก ์คํ
inputs = {'image': image_array}
outputs = session.run(None, inputs)
# ๊ฒฐ๊ณผ ํด์
results = {}
head_names = list(model_info['output_specification']['heads'].keys())
for i, output_name in enumerate(head_names):
logits = outputs[i]
probabilities = softmax(logits)[0]
# ํด๋์ค ์ด๋ฆ ๋งคํ
class_names = model_info['class_mappings'].get(output_name, {})
# ์ต๊ณ ํ๋ฅ ํด๋์ค
pred_idx = np.argmax(probabilities)
pred_class = class_names.get(str(pred_idx), f"Class_{pred_idx}")
pred_prob = probabilities[pred_idx]
# ์์ 3๊ฐ ํด๋์ค
top3_indices = np.argsort(probabilities)[-3:][::-1]
top3_results = []
for idx in top3_indices:
class_name = class_names.get(str(idx), f"Class_{idx}")
prob = probabilities[idx]
top3_results.append({'class': class_name, 'probability': float(prob)})
results[output_name] = {
'predicted_class': pred_class,
'confidence': float(pred_prob),
'top3': top3_results
}
return results
def predict_image_split_model(encoder_path, head_path, model_info_path, image_path):
"""๋ถ๋ฆฌ ๋ชจ๋ธ์ ์ฌ์ฉํ ์ด๋ฏธ์ง ๋ถ๋ฅ ์์ธก"""
# ๋ชจ๋ธ ์ ๋ณด ๋ก๋
model_info = load_model_info(model_info_path)
# ONNX ์ธ์
์์ฑ
encoder_session = ort.InferenceSession(encoder_path)
head_session = ort.InferenceSession(head_path)
# ์ด๋ฏธ์ง ์ ์ฒ๋ฆฌ
image_array = preprocess_image(image_path)
# ์ธ์ฝ๋๋ก ํน์ง ๋ฒกํฐ ์ถ์ถ
encoder_inputs = {'image': image_array}
features = encoder_session.run(None, encoder_inputs)[0]
# ํค๋๋ก ๋ถ๋ฅ
head_inputs = {'features': features}
outputs = head_session.run(None, head_inputs)
# ๊ฒฐ๊ณผ ํด์
results = {}
head_names = list(model_info['output_specification']['heads'].keys())
for i, output_name in enumerate(head_names):
logits = outputs[i]
probabilities = softmax(logits)[0]
# ํด๋์ค ์ด๋ฆ ๋งคํ
class_names = model_info['class_mappings'].get(output_name, {})
# ์ต๊ณ ํ๋ฅ ํด๋์ค
pred_idx = np.argmax(probabilities)
pred_class = class_names.get(str(pred_idx), f"Class_{pred_idx}")
pred_prob = probabilities[pred_idx]
# ์์ 3๊ฐ ํด๋์ค
top3_indices = np.argsort(probabilities)[-3:][::-1]
top3_results = []
for idx in top3_indices:
class_name = class_names.get(str(idx), f"Class_{idx}")
prob = probabilities[idx]
top3_results.append({'class': class_name, 'probability': float(prob)})
results[output_name] = {
'predicted_class': pred_class,
'confidence': float(pred_prob),
'top3': top3_results
}
return results
# ์ฌ์ฉ ์์
if __name__ == "__main__":
model_info_path = "model_info.json"
image_path = "test_image.jpg"
# ๋ถ๋ฆฌ ๋ชจ๋ธ์ด ์๋์ง ํ์ธ
if Path("encoder.onnx").exists() and Path("head.onnx").exists():
print("๋ถ๋ฆฌ ๋ชจ๋ธ ์ฌ์ฉ")
results = predict_image_split_model("encoder.onnx", "head.onnx", model_info_path, image_path)
elif Path("model.onnx").exists():
print("์ ์ฒด ๋ชจ๋ธ ์ฌ์ฉ")
results = predict_image_full_model("model.onnx", model_info_path, image_path)
else:
print("ONNX ๋ชจ๋ธ์ ์ฐพ์ ์ ์์ต๋๋ค.")
exit(1)
print(f"\n์ด๋ฏธ์ง ๋ถ๋ฅ ๊ฒฐ๊ณผ: {image_path}")
print("=" * 50)
for output_name, result in results.items():
print(f"\n{output_name.upper()}:")
print(f" ์์ธก ํด๋์ค: {result['predicted_class']}")
print(f" ์ ๋ขฐ๋: {result['confidence']:.4f}")
print(f" Top 3:")
for i, top_result in enumerate(result['top3'], 1):
print(f" {i}. {top_result['class']}: {top_result['probability']:.4f}")
|