Upload app.py

app.py

@@ -1,19 +1,46 @@
+import math
+
 import gradio as gr
 import tensorflow as tf
 
-new_model = tf.keras.models.load_model('my_model.h5')
-# preprocess_input = tf.keras.applications.resnet50.preprocess_input
+configs = [
+    {
+        "model": "my_model_2.h5", "size": 512
+    },
+    {
+        "model": "my_model.h5", "size": 224
+    },
+]
+
+config = configs[1]
+
+new_model = tf.keras.models.load_model(config["model"])
 
 def classify_image(inp):
-    inp = inp.reshape((-1, 224, 224, 3))
+    inp = inp.reshape((-1, config["size"], config["size"], 3))
     prediction = new_model.predict(inp).flatten()
-    if prediction[0] > 0:
-        return "Glaucoma"
-    return "Not glaucoma"
+    print(prediction)
+    print(new_model.predict(inp))
+    if len(prediction) > 1:
+        probability = 100 * math.exp(prediction[1]) / (math.exp(prediction[0]) + math.exp(prediction[1]))
+    else:
+        probability = round(100. / (1 + math.exp(-prediction[0])), 2)
+    if probability > 45:
+        return "Glaucoma", probability
+    if probability > 25:
+        return "Unclear", probability
+    return "Not glaucoma", probability
 
 
-gr.Interface(fn=classify_image, 
-             inputs=gr.inputs.Image(shape=(224, 224)),
-             outputs="label",
-             examples=[]).launch()
+gr.Interface(
+    fn=classify_image, 
+    inputs=gr.inputs.Image(shape=(224, 224)),
+    outputs=[
+    gr.outputs.Textbox(label="Label"),
+    gr.outputs.Textbox(label="Glaucoma probability (0 - 100)"),
+    ],
+    examples=["001.jpg", "002.jpg", "225.jpg"],
+    flagging_options=["Correct label", "Incorrect label"],
+    allow_flagging="manual",
+).launch()