upload code

flagged/log.csv

@@ -0,0 +1,2 @@
+noise_img,output,flag,username,timestamp
+C:\Users\qudwn\Desktop\prenet_web\flagged\noise_img\322f100a5a9a9da8a48e66336647bd6ae6509302\tmpoov2q1dh.jpg,C:\Users\qudwn\Desktop\prenet_web\flagged\output\771b11b68b939200ac2febaffce3f4c678885428\tmpfu7rv174.png,,,2023-10-16 22:43:42.103244

image_app.py

@@ -0,0 +1,58 @@
+import gradio as gr
+import torch
+from torch.autograd import Variable
+from utils import *
+from networks import *
+import cv2
+from torchvision import transforms
+import os
+
+model = PRN_r(6, 1)
+
+model = model.cuda()
+model.load_state_dict(torch.load('./net_latest.pth'))
+iscuda = torch.cuda.is_available()
+
+data_path = './datasets/Rain100H/'
+file_list = [data_path+file_name for file_name in os.listdir(data_path)]
+
+
+def deNosing(noise_img):
+
+    y = np.array(noise_img)
+    y = cv2.resize(y, (int(500), int(500)), interpolation=cv2.INTER_CUBIC)
+
+    y = normalize(np.float32(y))
+    y = np.expand_dims(y.transpose(2, 0, 1), 0)
+    y = Variable(torch.Tensor(y))
+
+    if iscuda:
+        y = y.cuda()
+
+    with torch.no_grad():
+        if iscuda:
+            torch.cuda.synchronize()
+
+        out, _ = model(y)
+        out = torch.clamp(out, 0., 1.)
+
+        if iscuda:
+            torch.cuda.synchronize()
+
+
+    if iscuda:
+        save_out = np.uint8(255 * out.data.cpu().numpy().squeeze())   #back to cpu
+    else:
+        save_out = np.uint8(255 * out.data.numpy().squeeze())
+
+    save_out = save_out.transpose(1, 2, 0)
+    save_out = cv2.resize(save_out,(noise_img.width,noise_img.height),interpolation=cv2.INTER_CUBIC)
+    return save_out
+
+demo = gr.Interface(deNosing,
+                    gr.inputs.Image(type='pil'),
+                    gr.outputs.Image(type='pil'),
+                    examples=file_list
+                    )
+
+demo.launch(debug=True, share=True)