Upload 3 files
Browse files- app.py +54 -0
- best_model_swin.pth +3 -0
- requirements.txt +5 -0
app.py
ADDED
|
@@ -0,0 +1,54 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# app.py (Use this code for Hugging Face)
|
| 2 |
+
import torch
|
| 3 |
+
import gradio as gr
|
| 4 |
+
from PIL import Image
|
| 5 |
+
from transformers import SwinForImageClassification, ViTImageProcessor
|
| 6 |
+
|
| 7 |
+
# --- 1. Load Model & Processor ---
|
| 8 |
+
MODEL_NAME = "microsoft/swin-tiny-patch4-window7-224"
|
| 9 |
+
MODEL_PATH = "best_model_swin.pth"
|
| 10 |
+
NUM_CLASSES = 3
|
| 11 |
+
CLASS_NAMES = ['COVID19', 'NORMAL', 'PNEUMONIA']
|
| 12 |
+
device = torch.device("cpu") # Use CPU for free-tier hosting
|
| 13 |
+
|
| 14 |
+
processor = ViTImageProcessor.from_pretrained(MODEL_NAME)
|
| 15 |
+
model = SwinForImageClassification.from_pretrained(
|
| 16 |
+
MODEL_NAME,
|
| 17 |
+
num_labels=NUM_CLASSES,
|
| 18 |
+
ignore_mismatched_sizes=True
|
| 19 |
+
)
|
| 20 |
+
model.load_state_dict(torch.load(MODEL_PATH, map_location=device))
|
| 21 |
+
model.to(device)
|
| 22 |
+
model.eval()
|
| 23 |
+
|
| 24 |
+
# --- 2. Define Prediction Function ---
|
| 25 |
+
def classify_image(input_image: Image.Image):
|
| 26 |
+
if input_image is None:
|
| 27 |
+
return "Please upload an image."
|
| 28 |
+
if input_image.mode != "RGB":
|
| 29 |
+
input_image = input_image.convert("RGB")
|
| 30 |
+
|
| 31 |
+
inputs = processor(images=input_image, return_tensors="pt")
|
| 32 |
+
pixel_values = inputs['pixel_values'].to(device)
|
| 33 |
+
|
| 34 |
+
with torch.no_grad():
|
| 35 |
+
outputs = model(pixel_values)
|
| 36 |
+
|
| 37 |
+
probabilities = torch.nn.functional.softmax(outputs.logits, dim=1)
|
| 38 |
+
|
| 39 |
+
# Create a dictionary of {class_name: probability}
|
| 40 |
+
confidences = {CLASS_NAMES[i]: prob.item() for i, prob in enumerate(probabilities[0])}
|
| 41 |
+
|
| 42 |
+
return confidences
|
| 43 |
+
|
| 44 |
+
# --- 3. Create the Gradio Interface ---
|
| 45 |
+
iface = gr.Interface(
|
| 46 |
+
fn=classify_image,
|
| 47 |
+
inputs=gr.Image(type="pil", label="Upload Chest X-Ray"),
|
| 48 |
+
outputs=gr.Label(num_top_classes=3, label="Predictions"),
|
| 49 |
+
title="Swin Transformer Chest X-Ray Classifier",
|
| 50 |
+
description="Upload an X-ray image to classify it as COVID-19, Normal, or Pneumonia."
|
| 51 |
+
)
|
| 52 |
+
|
| 53 |
+
# --- 4. Launch the app ---
|
| 54 |
+
iface.launch()
|
best_model_swin.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2ac3270d49c78130820f0ef852c827788dafe8780a19abd55deb3f807e4f3a9b
|
| 3 |
+
size 110416869
|
requirements.txt
ADDED
|
@@ -0,0 +1,5 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# requirements.txt
|
| 2 |
+
torch
|
| 3 |
+
transformers
|
| 4 |
+
gradio
|
| 5 |
+
pillow
|