add mpr predictor and doc (PaddlePaddle#530)

* add mpr predictor and doc * add doc * fix result * fix Results * fix copyright
xiedongmingming · Dec 21, 2021 · 2b28df6 · 2b28df6
1 parent 49ef722
commit 2b28df6
Show file tree

Hide file tree

Showing 15 changed files with 571 additions and 55 deletions.
diff --git a/applications/tools/animeganv2.py b/applications/tools/animeganv2.py
@@ -1,6 +1,21 @@
+#  Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserve.
+#
+#Licensed under the Apache License, Version 2.0 (the "License");
+#you may not use this file except in compliance with the License.
+#You may obtain a copy of the License at
+#
+#    http://www.apache.org/licenses/LICENSE-2.0
+#
+#Unless required by applicable law or agreed to in writing, software
+#distributed under the License is distributed on an "AS IS" BASIS,
+#WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+#See the License for the specific language governing permissions and
+#limitations under the License.
+
 import paddle
 import os
 import sys
+
 sys.path.insert(0, os.getcwd())
 from ppgan.apps import AnimeGANPredictor
 import argparse

diff --git a/applications/tools/face_parse.py b/applications/tools/face_parse.py
@@ -1,3 +1,17 @@
+#  Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserve.
+#
+#Licensed under the Apache License, Version 2.0 (the "License");
+#you may not use this file except in compliance with the License.
+#You may obtain a copy of the License at
+#
+#    http://www.apache.org/licenses/LICENSE-2.0
+#
+#Unless required by applicable law or agreed to in writing, software
+#distributed under the License is distributed on an "AS IS" BASIS,
+#WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+#See the License for the specific language governing permissions and
+#limitations under the License.
+
 import argparse
 
 import paddle

diff --git a/applications/tools/image_restoration.py b/applications/tools/image_restoration.py
@@ -1,6 +1,21 @@
+#  Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserve.
+#
+#Licensed under the Apache License, Version 2.0 (the "License");
+#you may not use this file except in compliance with the License.
+#You may obtain a copy of the License at
+#
+#    http://www.apache.org/licenses/LICENSE-2.0
+#
+#Unless required by applicable law or agreed to in writing, software
+#distributed under the License is distributed on an "AS IS" BASIS,
+#WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+#See the License for the specific language governing permissions and
+#limitations under the License.
+
 import paddle
 import os
 import sys
+
 sys.path.insert(0, os.getcwd())
 from ppgan.apps import MPRPredictor
 import argparse
@@ -22,18 +37,18 @@
                         default=None,
                         help="sample random seed for model's image generation")
 
-    parser.add_argument('--images_path', 
-                        default=None, 
+    parser.add_argument('--images_path',
+                        default=None,
                         required=True,
-                        type=str, 
+                        type=str,
                         help='Single image or images directory.')
 
-    parser.add_argument('--task', 
-                        required=True, 
-                        type=str, 
-                        help='Task to run', 
+    parser.add_argument('--task',
+                        required=True,
+                        type=str,
+                        help='Task to run',
                         choices=['Deblurring', 'Denoising', 'Deraining'])
-                        
+
     parser.add_argument("--cpu",
                         dest="cpu",
                         action="store_true",
@@ -44,11 +59,9 @@
     if args.cpu:
         paddle.set_device('cpu')
 
-    predictor = MPRPredictor(
-        images_path=args.images_path,
-        output_path=args.output_path,
-        weight_path=args.weight_path,
-        seed=args.seed,
-        task=args.task
-    )
+    predictor = MPRPredictor(images_path=args.images_path,
+                             output_path=args.output_path,
+                             weight_path=args.weight_path,
+                             seed=args.seed,
+                             task=args.task)
     predictor.run()
diff --git a/applications/tools/lapstyle.py b/applications/tools/lapstyle.py
@@ -1,3 +1,17 @@
+#  Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserve.
+#
+#Licensed under the Apache License, Version 2.0 (the "License");
+#you may not use this file except in compliance with the License.
+#You may obtain a copy of the License at
+#
+#    http://www.apache.org/licenses/LICENSE-2.0
+#
+#Unless required by applicable law or agreed to in writing, software
+#distributed under the License is distributed on an "AS IS" BASIS,
+#WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+#See the License for the specific language governing permissions and
+#limitations under the License.
+
 import paddle
 import os
 import sys

diff --git a/applications/tools/mprnet.py b/applications/tools/mprnet.py
@@ -0,0 +1,56 @@
+#  Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserve.
+#
+#Licensed under the Apache License, Version 2.0 (the "License");
+#you may not use this file except in compliance with the License.
+#You may obtain a copy of the License at
+#
+#    http://www.apache.org/licenses/LICENSE-2.0
+#
+#Unless required by applicable law or agreed to in writing, software
+#distributed under the License is distributed on an "AS IS" BASIS,
+#WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+#See the License for the specific language governing permissions and
+#limitations under the License.
+
+import paddle
+import os
+import sys
+
+sys.path.insert(0, os.getcwd())
+from ppgan.apps import MPRPredictor
+import argparse
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--input_image", type=str, help="path to image")
+
+    parser.add_argument("--output_path",
+                        type=str,
+                        default='output_dir',
+                        help="path to output image dir")
+
+    parser.add_argument("--weight_path",
+                        type=str,
+                        default=None,
+                        help="path to model weight path")
+
+    parser.add_argument(
+        "--task",
+        type=str,
+        default='Deblurring',
+        help="task can be chosen in 'Deblurring', 'Denoising', 'Deraining'")
+
+    parser.add_argument("--cpu",
+                        dest="cpu",
+                        action="store_true",
+                        help="cpu mode.")
+
+    args = parser.parse_args()
+
+    if args.cpu:
+        paddle.set_device('cpu')
+
+    predictor = MPRPredictor(output_path=args.output_path,
+                             task=args.task,
+                             weight_path=args.weight_path)
+    predictor.run(args.input_image)
diff --git a/applications/tools/photopen.py b/applications/tools/photopen.py
@@ -1,3 +1,17 @@
+#  Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserve.
+#
+#Licensed under the Apache License, Version 2.0 (the "License");
+#you may not use this file except in compliance with the License.
+#You may obtain a copy of the License at
+#
+#    http://www.apache.org/licenses/LICENSE-2.0
+#
+#Unless required by applicable law or agreed to in writing, software
+#distributed under the License is distributed on an "AS IS" BASIS,
+#WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+#See the License for the specific language governing permissions and
+#limitations under the License.
+
 import paddle
 import os
 import sys
@@ -39,10 +53,9 @@
 
     if args.cpu:
         paddle.set_device('cpu')
-        
+
     cfg = get_config(args.config_file)
     predictor = PhotoPenPredictor(output_path=args.output_path,
-                        weight_path=args.weight_path,
-                        gen_cfg=cfg.predict)
+                                  weight_path=args.weight_path,
+                                  gen_cfg=cfg.predict)
     predictor.run(semantic_label_path=args.semantic_label_path)
-
diff --git a/applications/tools/wav2lip.py b/applications/tools/wav2lip.py
@@ -1,3 +1,17 @@
+#  Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserve.
+#
+#Licensed under the Apache License, Version 2.0 (the "License");
+#you may not use this file except in compliance with the License.
+#You may obtain a copy of the License at
+#
+#    http://www.apache.org/licenses/LICENSE-2.0
+#
+#Unless required by applicable law or agreed to in writing, software
+#distributed under the License is distributed on an "AS IS" BASIS,
+#WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+#See the License for the specific language governing permissions and
+#limitations under the License.
+
 import argparse
 
 import paddle
@@ -103,29 +117,28 @@
     type=str,
     default='sfd',
     help="face detector to be used, can choose s3fd or blazeface")
-parser.add_argument(
-    "--face_enhancement",
-    dest="face_enhancement",
-    action="store_true",
-    help="use face enhance for face")
+parser.add_argument("--face_enhancement",
+                    dest="face_enhancement",
+                    action="store_true",
+                    help="use face enhance for face")
 parser.set_defaults(face_enhancement=False)
 
 if __name__ == "__main__":
     args = parser.parse_args()
     if args.cpu:
         paddle.set_device('cpu')
 
-    predictor = Wav2LipPredictor(checkpoint_path = args.checkpoint_path,
-                                 static = args.static,
-                                 fps = args.fps,
-                                 pads = args.pads,
-                                 face_det_batch_size = args.face_det_batch_size,
-                                 wav2lip_batch_size = args.wav2lip_batch_size,
-                                 resize_factor = args.resize_factor,
-                                 crop = args.crop,
-                                 box = args.box,
-                                 rotate = args.rotate,
-                                 nosmooth = args.nosmooth,
-                                 face_detector = args.face_detector,
-                                 face_enhancement = args.face_enhancement)
+    predictor = Wav2LipPredictor(checkpoint_path=args.checkpoint_path,
+                                 static=args.static,
+                                 fps=args.fps,
+                                 pads=args.pads,
+                                 face_det_batch_size=args.face_det_batch_size,
+                                 wav2lip_batch_size=args.wav2lip_batch_size,
+                                 resize_factor=args.resize_factor,
+                                 crop=args.crop,
+                                 box=args.box,
+                                 rotate=args.rotate,
+                                 nosmooth=args.nosmooth,
+                                 face_detector=args.face_detector,
+                                 face_enhancement=args.face_enhancement)
     predictor.run(args.face, args.audio, args.outfile)
diff --git a/configs/mprnet_deblurring.yaml b/configs/mprnet_deblurring.yaml
@@ -6,6 +6,9 @@ model:
   name: MPRModel
   generator:
     name: MPRNet
+    n_feat: 96
+    scale_unetfeats: 48
+    scale_orsnetfeats: 32
 
   char_criterion:
     name: CharbonnierLoss
@@ -15,16 +18,16 @@ model:
 dataset:
   train:
     name: MPRTrain
-    rgb_dir: 'data/GoPro/train'
+    rgb_dir: data/GoPro/train
     num_workers: 4
-    batch_size: 2
+    batch_size: 2 # 8GPUs
     img_options:
       patch_size: 256
   test:
     name: MPRVal
-    rgb_dir: 'data/GoPro/test'
-    num_workers: 4
-    batch_size: 2
+    rgb_dir: data/GoPro/test
+    num_workers: 1
+    batch_size: 1
     img_options:
       patch_size: 256
 

diff --git a/configs/mprnet_denoising.yaml b/configs/mprnet_denoising.yaml
@@ -0,0 +1,69 @@
+total_iters: 100000
+output_dir: output_dir
+
+model:
+  name: MPRModel
+  generator:
+    name: MPRNet
+    n_feat: 80
+    scale_unetfeats: 48
+    scale_orsnetfeats: 32
+
+  char_criterion:
+    name: CharbonnierLoss
+  edge_criterion:
+    name: EdgeLoss
+
+dataset:
+  train:
+    name: MPRTrain
+    rgb_dir: data/SIDD/train
+    num_workers: 16
+    batch_size: 4 # 4GPUs
+    img_options:
+      patch_size: 256
+  test:
+    name: MPRTrain
+    rgb_dir: data/SIDD/val
+    num_workers: 1
+    batch_size: 1
+    img_options:
+      patch_size: 256
+
+lr_scheduler:
+  name: CosineAnnealingRestartLR
+  learning_rate: !!float 2e-4
+  periods: [25000, 25000, 25000, 25000]
+  restart_weights: [1, 1, 1, 1]
+  eta_min: !!float 1e-6
+
+validate:
+  interval: 5000
+  save_img: false
+
+  metrics:
+    psnr: # metric name, can be arbitrary
+      name: PSNR
+      crop_border: 4
+      test_y_channel: True
+    ssim:
+      name: SSIM
+      crop_border: 4
+      test_y_channel: True
+
+optimizer:
+  name: Adam
+  # add parameters of net_name to optim
+  # name should in self.nets
+  net_names:
+    - generator
+  beta1: 0.9
+  beta2: 0.999
+  epsilon: 1e-8
+
+log_config:
+  interval: 10
+  visiual_interval: 5000
+
+snapshot_config:
+  interval: 5000