Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
|
@@ -3,68 +3,13 @@ os.system('pip install tf_keras')
|
|
| 3 |
|
| 4 |
import torch
|
| 5 |
from torchvision import transforms
|
| 6 |
-
|
| 7 |
import tensorflow as tf
|
| 8 |
-
import tensorflow_hub as hub
|
| 9 |
import tf_keras
|
| 10 |
-
from tf_keras.layers import Input, Dense, Dropout, GlobalAveragePooling2D
|
| 11 |
-
from tf_keras.models import Model, Sequential
|
| 12 |
-
from tf_keras.optimizers import Adam
|
| 13 |
-
from tf_keras.preprocessing import image
|
| 14 |
-
from tf_keras.applications.vgg16 import preprocess_input
|
| 15 |
-
|
| 16 |
import numpy as np
|
| 17 |
-
# from tensorflow.keras.layers import Input, Dense, Dropout, GlobalAveragePooling2D
|
| 18 |
-
# from tensorflow.keras.models import Model, Sequential
|
| 19 |
-
# from tensorflow.keras.optimizers import Adam
|
| 20 |
-
# from tensorflow.keras.preprocessing import image
|
| 21 |
-
# from tensorflow.keras.applications.vgg16 import preprocess_input
|
| 22 |
-
# from transformers import TFAutoModelForSequenceClassification
|
| 23 |
-
|
| 24 |
import gradio as gr
|
| 25 |
|
| 26 |
-
|
| 27 |
-
|
| 28 |
-
model_h5_path = "./model.h5"
|
| 29 |
-
|
| 30 |
-
# tf.config.experimental._could_use_gpu_kernel = False
|
| 31 |
-
|
| 32 |
-
# model = TFAutoModelForSequenceClassification.from_pretrained(model_pb_path)
|
| 33 |
-
model = tf_keras.models.load_model(model_pb_path)
|
| 34 |
-
# model = tf.keras.models.load_model(model_keras_path)
|
| 35 |
-
# model = tf.keras.models.load_model(model_h5_path)
|
| 36 |
-
# model.compile(optimizer=Adam(learning_rate=0.0001),
|
| 37 |
-
# loss='binary_crossentropy',
|
| 38 |
-
# metrics=['accuracy'])
|
| 39 |
-
|
| 40 |
-
#------------------------------------------------------------
|
| 41 |
-
# model_url = "https://tfhub.dev/sayakpaul/vit_b16_classification/1"
|
| 42 |
-
|
| 43 |
-
# # Load the ViT-B16 model pre-trained on ImageNet
|
| 44 |
-
# try:
|
| 45 |
-
# vit_b16 = hub.KerasLayer(model_url, trainable=False, input_shape=(224, 224, 3))
|
| 46 |
-
# except Exception as e:
|
| 47 |
-
# print(f"Failed to load the model from {model_url}. Error: {e}")
|
| 48 |
-
# # Handle the error if the model cannot be loaded
|
| 49 |
-
# raise
|
| 50 |
-
|
| 51 |
-
# # Create a Sequential model
|
| 52 |
-
# model = Sequential([
|
| 53 |
-
# vit_b16,
|
| 54 |
-
# Dense(1024, activation='relu'),
|
| 55 |
-
# Dropout(0.5),
|
| 56 |
-
# Dense(1, activation='sigmoid')
|
| 57 |
-
# ])
|
| 58 |
-
|
| 59 |
-
# # Compile the model
|
| 60 |
-
# model.compile(optimizer=Adam(learning_rate=0.0001),
|
| 61 |
-
# loss='binary_crossentropy',
|
| 62 |
-
# metrics=['accuracy'])
|
| 63 |
-
|
| 64 |
-
# weights_path = "./best_model_weights.h5"
|
| 65 |
-
# model.load_weights(weights_path)
|
| 66 |
-
#------------------------------------------------------------
|
| 67 |
-
|
| 68 |
|
| 69 |
# Define preprocessing function
|
| 70 |
def preprocess_image(image):
|
|
@@ -90,15 +35,28 @@ def predict(image):
|
|
| 90 |
predictions = model.predict(input_tensor)
|
| 91 |
|
| 92 |
# Extract the predicted class label with the highest probability
|
| 93 |
-
|
| 94 |
-
|
| 95 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
| 96 |
|
| 97 |
image_component = gr.Image(type="pil", label="Chest X-ray Image")
|
| 98 |
label_component = gr.Label(label="Classification result", num_top_classes=1)
|
| 99 |
description = ""
|
| 100 |
title = "Pneumonia Classifier"
|
| 101 |
|
| 102 |
-
demo = gr.Interface(fn=predict,
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 103 |
|
| 104 |
demo.launch()
|
|
|
|
| 3 |
|
| 4 |
import torch
|
| 5 |
from torchvision import transforms
|
|
|
|
| 6 |
import tensorflow as tf
|
|
|
|
| 7 |
import tf_keras
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 8 |
import numpy as np
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 9 |
import gradio as gr
|
| 10 |
|
| 11 |
+
model_path = "./4-15_ViT-B16_Lung3_v1/"
|
| 12 |
+
model = tf_keras.models.load_model(model_path)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 13 |
|
| 14 |
# Define preprocessing function
|
| 15 |
def preprocess_image(image):
|
|
|
|
| 35 |
predictions = model.predict(input_tensor)
|
| 36 |
|
| 37 |
# Extract the predicted class label with the highest probability
|
| 38 |
+
prediction = np.argmax(predictions[0])
|
| 39 |
+
predicted_classes = (prediction > 0.5).astype(int) # 0.5 threshold
|
| 40 |
+
|
| 41 |
+
if(prediction > 0.5):
|
| 42 |
+
return "Pneumonia"
|
| 43 |
+
else:
|
| 44 |
+
return "Normal"
|
| 45 |
|
| 46 |
image_component = gr.Image(type="pil", label="Chest X-ray Image")
|
| 47 |
label_component = gr.Label(label="Classification result", num_top_classes=1)
|
| 48 |
description = ""
|
| 49 |
title = "Pneumonia Classifier"
|
| 50 |
|
| 51 |
+
demo = gr.Interface(fn=predict,
|
| 52 |
+
inputs=image_component,
|
| 53 |
+
outputs=label_component,
|
| 54 |
+
description=description,
|
| 55 |
+
title=title,
|
| 56 |
+
examples=["demo_example_imgs/NORMAL-0001-0001.jpeg",
|
| 57 |
+
"demo_example_imgs/NORMAL2-IM-0337-0001.jpeg",
|
| 58 |
+
"demo_example_imgs/PNEUMONIA_bacteria_434.jpeg",
|
| 59 |
+
"demo_example_imgs/PNEUMONIA_virus_140.jpeg"
|
| 60 |
+
])
|
| 61 |
|
| 62 |
demo.launch()
|