Create app.py

app.py

@@ -0,0 +1,51 @@
+import os
+import copy
+import time
+
+import cv2 as cv
+import numpy as np
+import onnxruntime
+
+import gradio
+
+def run_inference(onnx_session, input_size, image):
+    # Pre process:Resize, BGR->RGB, Transpose, PyTorch standardization, float32 cast
+    temp_image = copy.deepcopy(image)
+    resize_image = cv.resize(temp_image, dsize=(input_size[0], input_size[1]))
+    x = cv.cvtColor(resize_image, cv.COLOR_BGR2RGB)
+    x = np.array(x, dtype=np.float32)
+    mean = [0.485, 0.456, 0.406]
+    std = [0.229, 0.224, 0.225]
+    x = (x / 255 - mean) / std
+    x = x.reshape(-1, input_size[0], input_size[1], 3).astype('float32')
+
+    # Inference
+    input_name = onnx_session.get_inputs()[0].name
+    output_name = onnx_session.get_outputs()[0].name
+    onnx_result = onnx_session.run([output_name], {input_name: x})
+
+    # Post process
+    onnx_result = np.array(onnx_result).squeeze()
+    onnx_result = (1 - onnx_result)
+    min_value = np.min(onnx_result)
+    max_value = np.max(onnx_result)
+    onnx_result = (onnx_result - min_value) / (max_value - min_value)
+    onnx_result *= 255
+    onnx_result = onnx_result.astype('uint8')
+
+    return onnx_result
+
+# Load model
+onnx_session = onnxruntime.InferenceSession(model_path)
+
+def create_rgba(image):
+  return run_inference(
+      onnx_session,
+      image.shape,
+      image,
+  )
+
+css = ".output_image {height: 100% !important; width: 100% !important;}"
+inputs = gradio.inputs.Image()
+outputs = gradio.outputs.Image()
+gradio.Interface(fn=create_rgba, inputs=inputs, outputs=outputs, css=css).launch()