updates files

app.py

@@ -1,6 +1,7 @@
 import tensorflow as tf
 import numpy as np
 import gradio as gr
+import matplotlib.pyplot as plt
 from huggingface_hub import from_pretrained_keras
 import os
 import sys
@@ -8,17 +9,31 @@ import sys
 print('Loading model...')
 model = from_pretrained_keras("mostafapasha/ribs-segmentation-model", compile=False)
 print('Successfully loaded model...')
-examples = ['VinDr_RibCXR_train_008.png', 'VinDr_RibCXR_train_013.png']
+examples = ['examples/VinDr_RibCXR_train_008.png', 'examples/VinDr_RibCXR_train_013.png']
 
 
-def segment(img_arr):
-    if np.ndim(img_arr) != 2:
-        img_arr = img_arr[:, :, 1]
-    image = img_arr[np.newaxis, :, :, np.newaxis]
-    threshold = 0.5
-    logits = model(image)
+def infer(img, threshold):
+    if np.ndim(img) != 2:
+        img = img[:, :, 1]
+    img = img.reshape(1, img.shape[0], img.shape[1], 1)
+    logits = model(img, training=False)
     prob = tf.sigmoid(logits)
     pred = tf.cast(prob > threshold, dtype=tf.float32)
     pred = np.array(pred.numpy())[0,:,:,0]
     return pred
-iface = gr.Interface(fn=segment, inputs=gr.inputs.Image(shape=(512, 512)), outputs="image", examples=examples, flagging_dir="flagged").launch(cache_examples=True)
+
+gr_input = [
+    gr.inputs.Image(label="Image", type="numpy", shape=(512, 512))
+    ,gr.inputs.Slider(minimum=0, maximum=1, step=0.05, default=0.5, label="Segmentation Threshold")
+]
+
+gr_output = [
+    gr.outputs.Image(type="pil",label="Segmentation Mask"),
+    # gr.outputs.Image(type="pil",label="Filtered Image"),
+]
+
+iface = gr.Interface(fn=infer,
+title = 'ribs segmentation model',
+description = 'Keras implementation of ResUNET++ for xray ribs segmentation',
+inputs=gr_input,
+outputs=gr_output, examples=examples, flagging_dir="flagged").launch(cache_examples=True)

requirements.txt

@@ -1,2 +1,6 @@
-tensorflow
-keras
+gradio==2.9.4
+huggingface_hub==0.6.0
+matplotlib==3.5.1
+numpy==1.22.3
+tensorflow==2.9.0
+tensorflow_macos==2.8.0