Upload 7 files

.gitattributes

@@ -33,3 +33,8 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+images/00_00.png filter=lfs diff=lfs merge=lfs -text
+images/01_00.png filter=lfs diff=lfs merge=lfs -text
+images/02_00.png filter=lfs diff=lfs merge=lfs -text
+images/result.png filter=lfs diff=lfs merge=lfs -text
+model/codeformer.axmodel filter=lfs diff=lfs merge=lfs -text

model/codeformer.axmodel

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fd97c1ec3b2fb0927c375c2b3d8e149137fa34e668e6d164654e3933a8d9fd88
+size 94975113

python/requirements.txt

@@ -0,0 +1,2 @@
+numpy
+opencv-python

python/run_axmodel.py

@@ -0,0 +1,65 @@
+import argparse
+import os
+import cv2
+import numpy as np
+import axengine as axe
+
+def from_numpy(x):
+    return x if isinstance(x, np.ndarray) else np.array(x)
+
+def main(args):
+    # Initialize the model
+    session = axe.InferenceSession(args.model_path)
+    output_names = [x.name for x in session.get_outputs()]
+    input_name = session.get_inputs()[0].name
+
+    # results
+    os.makedirs(args.output_path, exist_ok=True)
+
+    files =[f for f in os.listdir(args.inputs_path) if f.lower().endswith(('.jpg', '.png', 'jpeg'))]
+    
+    for file in files:
+        ori_image = cv2.imread(os.path.join(args.inputs_path, file))
+        h, w = ori_image.shape[:2]
+        image = cv2.resize(ori_image, (512, 512))
+        image = (image[..., ::-1] /255.0).astype(np.float32)
+        
+        mean = [0.5, 0.5, 0.5]
+        std = [0.5, 0.5, 0.5]
+        image = ((image - mean) / std).astype(np.float32)
+
+        #image = (image /1.0).astype(np.float32)
+        img = np.transpose(np.expand_dims(np.ascontiguousarray(image), axis=0), (0,3,1,2))
+        
+        # Use the model to generate super-resolved images
+        sr = session.run(output_names, {input_name: img})
+
+        #sr_y_image = imgproc.array_to_image(sr)
+        sr = np.transpose(sr[0].squeeze(0), (1,2,0))
+        sr = (sr*std + mean).astype(np.float32)
+        
+        # Save image
+        ndarr = np.clip((sr*255.0), 0, 255.0).astype(np.uint8)
+        out_image = cv2.resize(ndarr[..., ::-1], (w, h))
+
+        cv2.imwrite(f'{arg.output_path}/{file}', out_image)
+        print(f"SR image save to `{file}`")
+
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser(description="Using the model generator super-resolution images.")
+    parser.add_argument("--inputs_path",
+                        type=str,
+                        default="images",
+                        help="origin image path.")
+    parser.add_argument("--output_path",
+                        type=str,
+                        default="results",
+                        help="colorized image path.")
+    parser.add_argument("--model_path",
+                        type=str,
+                        default="./codeformer.axmoel",
+                        help="model path.")
+    args = parser.parse_args()
+
+    main(args)