Update app.py
Browse files
app.py
CHANGED
@@ -1,3 +1,50 @@
|
|
1 |
-
import
|
|
|
|
|
|
|
|
|
|
|
|
|
2 |
|
3 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
from fastapi import FastAPI, UploadFile, File
|
2 |
+
from transformers import ViTForImageClassification, ViTFeatureExtractor
|
3 |
+
import torch
|
4 |
+
import torch.nn as nn
|
5 |
+
import torchvision.transforms as transforms
|
6 |
+
from PIL import Image
|
7 |
+
import io
|
8 |
|
9 |
+
app = FastAPI()
|
10 |
+
|
11 |
+
# Load the ViT model and its feature extractor
|
12 |
+
model_name = "google/vit-base-patch16-224-in21k"
|
13 |
+
model = ViTForImageClassification.from_pretrained(model_name)
|
14 |
+
feature_extractor = ViTFeatureExtractor.from_pretrained(model_name)
|
15 |
+
|
16 |
+
# Load the trained model weights
|
17 |
+
num_classes = 7
|
18 |
+
model.classifier = nn.Linear(model.config.hidden_size, num_classes)
|
19 |
+
# Load the trained weights
|
20 |
+
model.load_state_dict(torch.load("models/Anwarkh1/Skin_Cancer-Image_Classification", map_location=torch.device('cpu')))
|
21 |
+
model.eval()
|
22 |
+
|
23 |
+
# Define class labels
|
24 |
+
class_labels = ['benign_keratosis-like_lesions', 'basal_cell_carcinoma', 'actinic_keratoses', 'vascular_lesions', 'melanocytic_Nevi', 'melanoma', 'dermatofibroma']
|
25 |
+
|
26 |
+
# Define image transformations
|
27 |
+
transform = transforms.Compose([
|
28 |
+
transforms.Resize((224, 224)),
|
29 |
+
transforms.ToTensor(),
|
30 |
+
])
|
31 |
+
|
32 |
+
# Define API endpoint for model inference
|
33 |
+
@app.post('/predict')
|
34 |
+
async def predict(file: UploadFile = File(...)):
|
35 |
+
contents = await file.read()
|
36 |
+
image = Image.open(io.BytesIO(contents))
|
37 |
+
image = transform(image).unsqueeze(0) # Add batch dimension
|
38 |
+
|
39 |
+
with torch.no_grad():
|
40 |
+
outputs = model(image)
|
41 |
+
|
42 |
+
# Calculate softmax probabilities
|
43 |
+
probabilities = torch.softmax(outputs.logits, dim=1)
|
44 |
+
|
45 |
+
# Get predicted class index and its probability
|
46 |
+
predicted_idx = torch.argmax(probabilities).item()
|
47 |
+
predicted_label = class_labels[predicted_idx]
|
48 |
+
predicted_accuracy = probabilities[0][predicted_idx].item()
|
49 |
+
|
50 |
+
return {'predicted_class': predicted_label, 'accuracy': predicted_accuracy}
|