Spaces:
Sleeping
Sleeping
MS-YUN
commited on
Commit
•
bc3b9d1
1
Parent(s):
85e92ff
트랜스포머"
Browse files
app.py
CHANGED
@@ -1,8 +1,35 @@
|
|
|
|
1 |
|
|
|
|
|
|
|
|
|
|
|
2 |
|
3 |
-
|
4 |
-
return {'cat': 0.3, 'dog': 0.7}
|
5 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
6 |
import gradio as gr
|
7 |
-
|
8 |
-
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# 모델로딩
|
2 |
|
3 |
+
# ImageNet-1k에 훈련된 모델과 특징 추출기 로드
|
4 |
+
from transformers import ViTImageProcessor, ViTForImageClassification
|
5 |
+
model_name = "google/vit-base-patch16-224"
|
6 |
+
model = ViTForImageClassification.from_pretrained(model_name)
|
7 |
+
image_processor = ViTImageProcessor.from_pretrained(model_name)
|
8 |
|
9 |
+
# 이미지 예측 분류함수
|
|
|
10 |
|
11 |
+
import torch
|
12 |
+
def classify_image(inp):
|
13 |
+
# 이미지를 특징 벡터로 변환
|
14 |
+
inputs = image_processor(images=inp, return_tensors="pt")
|
15 |
+
pixel_values = inputs["pixel_values"]
|
16 |
+
|
17 |
+
# 예측 수행
|
18 |
+
outputs = model(pixel_values)
|
19 |
+
logits = outputs.logits
|
20 |
+
predicted_index = torch.argmax(logits, 1)[0].item()
|
21 |
+
|
22 |
+
# 가장 확률이 높은 라벨 반환``
|
23 |
+
label = model.config.id2label[predicted_index]
|
24 |
+
return label
|
25 |
+
|
26 |
+
# Gradio 인터페이스 설정
|
27 |
+
from PIL import Image
|
28 |
import gradio as gr
|
29 |
+
interface = gr.Interface(
|
30 |
+
fn=classify_image,
|
31 |
+
inputs=gr.components.Image(type="pil", label="Upload an Image"),
|
32 |
+
outputs="text",
|
33 |
+
live=True
|
34 |
+
)
|
35 |
+
interface.launch()
|