app.py

import gradio as gr
import pathlib
import tensorflow as tf

current_dir = pathlib.Path(__file__).parent

# images = [str(current_dir / "cheetah1.jpeg"), str(current_dir / "cheetah1.jpg"), str(current_dir / "lion.jpg")]

images = [str(current_dir / "data/benign/benign_4.jpg"), str(current_dir / "data/benign/benign_5.jpg"), str(current_dir / "data/benign/benign_6.jpg"), str(current_dir / "data/malignant/malignant_4.jpg"), str(current_dir / "data/malignant/malignant_5.jpg"), str(current_dir / "data/malignant/malignant_6.jpg")]


# img_classifier = gr.Interface.load(
#     "models/google/vit-base-patch16-224", examples=images, cache_examples=False
# )


# def func(img, text):
#     return img_classifier(img), text


# using_img_classifier_as_function = gr.Interface(
#     func,
#     [gr.Image(type="filepath"), "text"],
#     ["label", "text"],
#     examples=[
#         [str(current_dir / "cheetah1.jpeg"), None],
#         [str(current_dir / "cheetah1.jpg"), "cheetah"],
#         [str(current_dir / "lion.jpg"), "lion"],
#     ],
#     cache_examples=False,
# )
# demo = gr.TabbedInterface([using_img_classifier_as_function, img_classifier])

# if __name__ == "__main__":
#     demo.launch()
















# import gradio as gr
from tensorflow import keras
from skimage.transform import resize

# def greet(name):
#     return "Hello " + name + "!!"

# iface = gr.Interface(fn=greet, inputs="text", outputs="text")
# iface.launch()

# oc_resnet50_model1 = keras.models.load_model('./models/oc_model.h5')
print("current_dir", current_dir)
oc_resnet50_model2 = keras.models.load_model(f"{current_dir}/models/mendeley_oc_model_v2.h5")
labels = ['Benign Lesion', 'Malignant Lesion']

def classify_image(inp):

  # inp =resize(inp, (300, 300, 3))
  inp = inp.reshape((-1, 300, 300, 3))
  # inp = tf.keras.applications.mobilenet_v2.preprocess_input(inp)
  inp = tf.keras.applications.resnet50.preprocess_input(inp)
  prediction = oc_resnet50_model2.predict(inp).flatten()
  confidences = {labels[i]: float(prediction[i]) for i in range(2)}
  return confidences

gr.Interface(fn=classify_image, 
             inputs=gr.Image(shape=(300, 300)),
             outputs=gr.Label(num_top_classes=2),
             examples=images, cache_examples=False,
            #  interpretation="shap", num_shap=5
             ).launch()