upd

inference.py

@@ -1,86 +1,86 @@
-import os, sys, argparse, warnings
+import os, sys, argparse, warnings
-import torch, torch.nn as nn, torch.nn.functional as F
+import torch, torch.nn as nn, torch.nn.functional as F
-import numpy as np, cv2
+import numpy as np, cv2
-from PIL import Image
+from PIL import Image
-from model import DocScanner
+from model import DocScanner
-from seg import U2NETP
+from seg import U2NETP
-
+
-warnings.filterwarnings('ignore')
+warnings.filterwarnings('ignore')
-
+
-# Запоминаем CWD пользователя ДО смены директории
+# Запоминаем CWD пользователя ДО смены директории
-USER_CWD = os.getcwd()
+USER_CWD = os.getcwd()
-SCRIPT_DIR = os.path.dirname(os.path.abspath(__file__))
+SCRIPT_DIR = os.path.dirname(os.path.abspath(__file__))
-os.chdir(SCRIPT_DIR)
+os.chdir(SCRIPT_DIR)
-
+
-
+
-class Net(nn.Module):
+class Net(nn.Module):
-    def __init__(self):
+    def __init__(self):
-        super().__init__()
+        super().__init__()
-        self.msk = U2NETP(3, 1)
+        self.msk = U2NETP(3, 1)
-        self.bm = DocScanner()
+        self.bm = DocScanner()
-
+
-    def forward(self, x):
+    def forward(self, x):
-        msk, *_ = self.msk(x)
+        msk, *_ = self.msk(x)
-        bm = self.bm((msk > 0.5).float() * x, iters=12, test_mode=True)
+        bm = self.bm((msk > 0.5).float() * x, iters=12, test_mode=True)
-        return (2 * (bm / 286.8) - 1) * 0.99
+        return (2 * (bm / 286.8) - 1) * 0.99
-
+
-
+
-def load_model(model, path, strip_prefix=False):
+def load_model(model, path, strip_prefix=False):
-    state_dict = model.state_dict()
+    state_dict = model.state_dict()
-    pretrained = torch.load(path, map_location='cuda:0')
+    pretrained = torch.load(path, map_location='cuda:0')
-    if strip_prefix:
+    if strip_prefix:
-        pretrained = {k[6:]: v for k, v in pretrained.items() if k[6:] in state_dict}
+        pretrained = {k[6:]: v for k, v in pretrained.items() if k[6:] in state_dict}
-    else:
+    else:
-        pretrained = {k: v for k, v in pretrained.items() if k in state_dict}
+        pretrained = {k: v for k, v in pretrained.items() if k in state_dict}
-    state_dict.update(pretrained)
+    state_dict.update(pretrained)
-    model.load_state_dict(state_dict)
+    model.load_state_dict(state_dict)
-    return model
+    return model
-
+
-
+
-def resolve_user_path(path):
+def resolve_user_path(path):
-    """Если путь относительный — считаем его относительно CWD пользователя, не скрипта"""
+    """Если путь относительный — считаем его относительно CWD пользователя, не скрипта"""
-    if os.path.isabs(path):
+    if os.path.isabs(path):
-        return path
+        return path
-    return os.path.join(USER_CWD, path)
+    return os.path.join(USER_CWD, path)
-
+
-
+
-def main():
+def main():
-    parser = argparse.ArgumentParser()
+    parser = argparse.ArgumentParser()
-    parser.add_argument('-i', '--input', required=True)
+    parser.add_argument('-i', '--input', required=True)
-    parser.add_argument('-o', '--output', required=True)
+    parser.add_argument('-o', '--output', required=True)
-    opt = parser.parse_args()
+    opt = parser.parse_args()
-
+
-    input_path = resolve_user_path(opt.input)
+    input_path = resolve_user_path(opt.input)
-    output_path = resolve_user_path(opt.output)
+    output_path = resolve_user_path(opt.output)
-
+
-    net = Net().cuda().eval()
+    net = Net().cuda().eval()
-    load_model(net.msk, f'{SCRIPT_DIR}/model_pretrained/seg.pth', strip_prefix=True)
+    load_model(net.msk, f'{SCRIPT_DIR}/model_pretrained/seg.pth', strip_prefix=True)
-    load_model(net.bm, f'{SCRIPT_DIR}/model_pretrained/DocScanner-L.pth')
+    load_model(net.bm, f'{SCRIPT_DIR}/model_pretrained/DocScanner-L.pth')
-
+
-    image = np.array(Image.open(input_path))[:, :, :3] / 255.0
+    image = np.array(Image.open(input_path))[:, :, :3] / 255.0
-    height, width = image.shape[:2]
+    height, width = image.shape[:2]
-    tensor = torch.from_numpy(cv2.resize(image, (288, 288)).transpose(2, 0, 1)).float().unsqueeze(0)
+    tensor = torch.from_numpy(cv2.resize(image, (288, 288)).transpose(2, 0, 1)).float().unsqueeze(0)
-
+
-    with torch.no_grad():
+    with torch.no_grad():
-        bm = net(tensor.cuda()).cpu()
+        bm = net(tensor.cuda()).cpu()
-
+
-    flow = torch.from_numpy(np.stack([
+    flow = torch.from_numpy(np.stack([
-        cv2.blur(cv2.resize(bm[0, 0].numpy(), (width, height)), (3, 3)),
+        cv2.blur(cv2.resize(bm[0, 0].numpy(), (width, height)), (3, 3)),
-        cv2.blur(cv2.resize(bm[0, 1].numpy(), (width, height)), (3, 3))
+        cv2.blur(cv2.resize(bm[0, 1].numpy(), (width, height)), (3, 3))
-    ], axis=2)).unsqueeze(0)
+    ], axis=2)).unsqueeze(0)
-
+
-    out = F.grid_sample(
+    out = F.grid_sample(
-        torch.from_numpy(image).permute(2, 0, 1).unsqueeze(0).float(),
+        torch.from_numpy(image).permute(2, 0, 1).unsqueeze(0).float(),
-        flow, align_corners=True
+        flow, align_corners=True
-    )
+    )
-    result = ((out[0] * 255).permute(1, 2, 0).numpy()[:, :, ::-1]).astype(np.uint8)
+    result = ((out[0] * 255).permute(1, 2, 0).numpy()[:, :, ::-1]).astype(np.uint8)
-
+
-    out_dir = os.path.dirname(output_path)
+    out_dir = os.path.dirname(output_path)
-    if out_dir:
+    if out_dir:
-        os.makedirs(out_dir, exist_ok=True)
+        os.makedirs(out_dir, exist_ok=True)
-    cv2.imwrite(output_path, result)
+    cv2.imwrite(output_path, result)
-    print(f"[OK] {output_path}")
+    print(f"[OK] {output_path}")
-
+
-
+
-if __name__ == "__main__":
+if __name__ == "__main__":
-    main()
+    main()