Upload 2 files
Browse files- app.py +54 -0
- requirements.txt +8 -0
app.py
ADDED
@@ -0,0 +1,54 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import gradio as gr
|
2 |
+
import torch
|
3 |
+
import torchvision.models as models
|
4 |
+
from huggingface_hub import hf_hub_download
|
5 |
+
import json
|
6 |
+
from PIL import Image
|
7 |
+
import torchvision.transforms as transforms
|
8 |
+
|
9 |
+
# Set device
|
10 |
+
device = "cuda" if torch.cuda.is_available() else "cpu"
|
11 |
+
|
12 |
+
# Load model weights and class labels
|
13 |
+
weights_path = hf_hub_download(repo_id="AventIQ-AI/resnet18-cataract-detection-system", filename="cataract_detection_resnet18_quantized.pth")
|
14 |
+
labels_path = hf_hub_download(repo_id="AventIQ-AI/resnet18-cataract-detection-system", filename="class_names.json")
|
15 |
+
|
16 |
+
with open(labels_path, "r") as f:
|
17 |
+
class_labels = json.load(f)
|
18 |
+
|
19 |
+
# Load model
|
20 |
+
model = models.resnet18(pretrained=False)
|
21 |
+
num_classes = len(class_labels)
|
22 |
+
model.fc = torch.nn.Linear(in_features=512, out_features=num_classes)
|
23 |
+
model.load_state_dict(torch.load(weights_path, map_location=device))
|
24 |
+
model.to(device)
|
25 |
+
model.eval()
|
26 |
+
|
27 |
+
# Define transform
|
28 |
+
transform = transforms.Compose([
|
29 |
+
transforms.Resize((224, 224)),
|
30 |
+
transforms.ToTensor(),
|
31 |
+
transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])
|
32 |
+
])
|
33 |
+
|
34 |
+
# Prediction function
|
35 |
+
def predict(image):
|
36 |
+
image = transform(image).unsqueeze(0).to(device) # Preprocess image
|
37 |
+
with torch.no_grad():
|
38 |
+
output = model(image)
|
39 |
+
_, predicted = torch.max(output, 1)
|
40 |
+
predicted_class = class_labels[predicted.item()]
|
41 |
+
return {predicted_class: 1.0} # Confidence is assumed to be 1.0 for simplicity
|
42 |
+
|
43 |
+
# Gradio Interface
|
44 |
+
demo = gr.Interface(
|
45 |
+
fn=predict,
|
46 |
+
inputs=gr.Image(type="pil", label="Upload an Eye Image"),
|
47 |
+
outputs=gr.Label(label="Cataract Detection Result"),
|
48 |
+
title="ποΈ Cataract Detection System π₯",
|
49 |
+
description="π¬ Upload an eye image, and the AI model will determine if cataract is present! π",
|
50 |
+
theme="huggingface",
|
51 |
+
allow_flagging="never",
|
52 |
+
)
|
53 |
+
|
54 |
+
demo.launch()
|
requirements.txt
ADDED
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
torch
|
2 |
+
transformers
|
3 |
+
gradio
|
4 |
+
sentencepiece
|
5 |
+
torchvision
|
6 |
+
huggingface_hub
|
7 |
+
pillow
|
8 |
+
numpy
|