Spaces:
Sleeping
Sleeping
Changed code to load TF model with tf.keras.models.load_model() and finished the rest implementation
Browse files
app.py
CHANGED
|
@@ -1,38 +1,44 @@
|
|
| 1 |
import gradio as gr
|
| 2 |
-
from transformers import TFAutoModel
|
|
|
|
| 3 |
|
| 4 |
-
|
| 5 |
-
model =
|
| 6 |
-
tokenizer = AutoTokenizer.from_pretrained(PATH)
|
| 7 |
|
| 8 |
# Define preprocessing function
|
| 9 |
def preprocess_image(image):
|
| 10 |
transform_test = transforms.Compose([
|
| 11 |
transforms.Resize((224, 224)),
|
|
|
|
|
|
|
| 12 |
])
|
| 13 |
|
| 14 |
-
|
| 15 |
-
return
|
| 16 |
|
| 17 |
# Define inference function
|
| 18 |
def predict(image):
|
| 19 |
# Preprocess the image
|
| 20 |
preprocessed_image = preprocess_image(image)
|
| 21 |
|
| 22 |
-
# Convert preprocessed image to input format expected by the model
|
| 23 |
-
inputs = tokenizer(preprocessed_image, return_tensors="tf")
|
| 24 |
|
| 25 |
# Perform model inference
|
| 26 |
-
outputs = model(**inputs)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 27 |
|
| 28 |
-
#
|
| 29 |
-
|
| 30 |
-
predicted_labels = np.argmax(outputs.logits.numpy(), axis=1)
|
| 31 |
|
| 32 |
-
|
| 33 |
-
return predicted_labels
|
| 34 |
|
| 35 |
-
image_component = gr.Image(type="
|
| 36 |
label_component = gr.Label(label="Classification result", num_top_classes=1)
|
| 37 |
description = ""
|
| 38 |
title = "Pneumonia Classifier"
|
|
|
|
| 1 |
import gradio as gr
|
| 2 |
+
# from transformers import TFAutoModel
|
| 3 |
+
import tensorflow as tf
|
| 4 |
|
| 5 |
+
model_path = "./4-15_ViT-B16_Lung3_v1"
|
| 6 |
+
model = tf.keras.models.load_model(model_path)
|
|
|
|
| 7 |
|
| 8 |
# Define preprocessing function
|
| 9 |
def preprocess_image(image):
|
| 10 |
transform_test = transforms.Compose([
|
| 11 |
transforms.Resize((224, 224)),
|
| 12 |
+
transforms.ToTensor(),
|
| 13 |
+
transforms.Normalize((0.485, 0.456, 0.406), (0.229, 0.224, 0.225)),
|
| 14 |
])
|
| 15 |
|
| 16 |
+
preprocessed_img = transform_test(image)
|
| 17 |
+
return preprocessed_img
|
| 18 |
|
| 19 |
# Define inference function
|
| 20 |
def predict(image):
|
| 21 |
# Preprocess the image
|
| 22 |
preprocessed_image = preprocess_image(image)
|
| 23 |
|
| 24 |
+
# # Convert preprocessed image to input format expected by the model
|
| 25 |
+
# inputs = tokenizer(preprocessed_image, return_tensors="tf")
|
| 26 |
|
| 27 |
# Perform model inference
|
| 28 |
+
# outputs = model(**inputs)
|
| 29 |
+
|
| 30 |
+
# Add batch dimension to input image
|
| 31 |
+
input_tensor = tf.expand_dims(preprocessed_image, axis=0)
|
| 32 |
+
|
| 33 |
+
# Perform inference
|
| 34 |
+
predictions = model.predict(input_tensor)
|
| 35 |
|
| 36 |
+
# Extract the predicted class label with the highest probability
|
| 37 |
+
predicted_label = np.argmax(predictions[0])
|
|
|
|
| 38 |
|
| 39 |
+
return predicted_label
|
|
|
|
| 40 |
|
| 41 |
+
image_component = gr.Image(type="piL", label="Chest X-ray Image")
|
| 42 |
label_component = gr.Label(label="Classification result", num_top_classes=1)
|
| 43 |
description = ""
|
| 44 |
title = "Pneumonia Classifier"
|