PR to merge

README.md

@@ -31,7 +31,7 @@ Install python package apex for distributed training
 ```
 git clone https://github.com/NVIDIA/apex
 cd apex
-pip install -v --disable-pip-version-check --no-cache-dir ./
+pip install -v --disable-pip-version-check --no-cache-dir --global-option="--cpp_ext" --global-option="--cuda_ext" ./
 ```
 
 ## Supported model weights
@@ -48,10 +48,29 @@ Move the downloaded weights to weights/
 
 ## Preparation of dataset - Cityscapes
 
-Please download the dataset from the officit site - [Download](https://www.cityscapes-dataset.com/)
+Please download the dataset from the official site. 
+
+This dataset requires you to download the source data manually:
+You have to download files from - [Download](https://www.cityscapes-dataset.com/) (This dataset requires registration). The config file is written for leftImg8bit_sequence_trainvaltest.zip and the fine annotations files from gtFine_trainvaltest.zip. To use other datasets, you require additional configa, which are not included.
+
+Aternatively, For downloading using command line, as shared by [cemsaz](https://github.com/cemsaz/city-scapes-script), 
+
+Use below cmd by specifying your username and password,
+
+```
+wget --keep-session-cookies --save-cookies=cookies.txt --post-data 'username=myusername&password=mypassword&submit=Login' https://www.cityscapes-dataset.com/login/
+```
+
+and provide the packageID of the required zip file. 
+
+```
+wget --load-cookies cookies.txt --content-disposition https://www.cityscapes-dataset.com/file-handling/?packageID=1
+```
+
+Hint : You can get the package id from the download link of the file you need to download. In our case, for leftImg8bit_sequence_trainvaltest.zip and gtFine_trainvaltest.zip, it is packageID=14 & 1. 
 
 ```
-data_path = './data/leftImg8bit_sequence_trainvaltest_2K/'
+data_path = './data/cityscapes'
 ```
 
 Modify the data_path in config/cityscapes.py
@@ -60,7 +79,7 @@ Modify the data_path in config/cityscapes.py
 
 ```
 CUDA_VISIBLE_DEVICES=0,1 python -m torch.distributed.launch --master_port 1111 \ 
---nproc_per_node 2 python main.py --segnet <segnet_name> --dataset <dataset_name> \
+--nproc_per_node 2 main.py --segnet <segnet_name> --dataset <dataset_name> \
 --optical-flow-network <of_name> --checkname <SAVE_DIR>
 ```
 

config/camvid.py

@@ -23,8 +23,8 @@ class camvid_config(object):
     bsnet_weight_path = './weights/cityscapes-bisenet-R18.pth'
 
 
-    #resume_path = './weights/gsvnet_bisenet_r18.tar'
-    resume_path = './weights/gsvnet_swnet_r18.tar'
+    bisenet_resume_path = './weights/gsvnet_bisenet_r18.tar'
+    swnet_resume_path = './weights/gsvnet_swnet_r18.tar'
 
     optical_flow_network_path = './weights/flownet.pth.tar'
-    data_path = '' #put your data path here
+    data_path = './data/camvid' #put your data path here

config/cityscapes.py

@@ -29,7 +29,7 @@ class cityscapes_config(object):
     ]  # 19 classes + 1 void class
     swnet_weight_path = './weights/cityscapes-swnet-R18.pt'
     bsnet_weight_path = './weights/cityscapes-bisenet-R18.pth'
-    #resume_path = './weights/gsvnet_bisenet_r18.tar'
-    resume_path = './weights/gsvnet_swnet_r18.tar'
+    bisenet_resume_path = './weights/gsvnet_bisenet_r18.tar'
+    swnet_resume_path = './weights/gsvnet_swnet_r18.tar'
     optical_flow_network_path = './weights/flownet.pth.tar'
-    data_path = '' # put your data path here
+    data_path = './data/cityscapes' # put your data path here

requirements.txt

@@ -6,6 +6,7 @@ numpy
 Pillow
 scikit-learn
 tensorboardX
-torch==1.1.0
-torchvision==0.2.1
+torch>=1.1.0
+torchvision>=0.2.1
 tqdm
+opencv-python

trainer/base.py

@@ -163,7 +163,7 @@ def __init__(self, args, name):
 
         if args.dataset.find("cityscapes") >= 0:
             self.dataset_config = cityscapes_config()
-        elif args.dataset_name == "camvid":
+        elif args.dataset == "camvid":
             self.dataset_config = camvid_config()
         else:
             raise NotImplementedError("Trainer dataset %s is not registered into the system" % args.dataset)
@@ -173,7 +173,8 @@ def __init__(self, args, name):
         self.num_classes = self.dataset_config.num_classes
         self.swnet_weight_path = self.dataset_config.swnet_weight_path
         self.bsnet_weight_path = self.dataset_config.bsnet_weight_path
-        self.resume_path = self.dataset_config.resume_path
+        self.swnet_resume_path = self.dataset_config.swnet_resume_path
+        self.bisenet_resume_path = self.dataset_config.bisenet_resume_path
         self.optical_flow_network_path = self.dataset_config.optical_flow_network_path
         self.data_path = self.dataset_config.data_path
 
@@ -526,4 +527,4 @@ def get_images_from_lines(self, path, lines, return_dict):
             for frame in range(gt_frame_num-4,gt_frame_num+1):
                 frame_name = line[:-22] + ( "%06d" % (frame) ) + line[-16:]
                 frame_path = os.path.join(path, frame_name)
-                return_dict[frame_path] = image_loader(frame_path)
+                return_dict[frame_path] = image_loader(frame_path)

trainer/gsvnet_trainer.py

@@ -62,13 +62,15 @@ def _img_input_transform(self, args):
 
     def _val_img_input_transform(self, args):
         if self.args.segnet == 'bisenet':
+            self.resume_path = self.bisenet_resume_path
             mean_std = ([0.406, 0.456, 0.485], [0.225, 0.224, 0.229])
             img_transform = standard_transforms.Compose([
                 FlipChannels(),
                 standard_transforms.ToTensor(),
                 standard_transforms.Normalize(*mean_std)
             ])
         elif self.args.segnet == 'swiftnet':
+            self.resume_path = self.swnet_resume_path
             mean_std = ([72.3, 82.90, 73.15],[47.73, 48.49, 47.67])
             img_transform = standard_transforms.Compose([
                 FlipChannels(),
@@ -478,4 +480,4 @@ def load_state(self, args):
                 self.test_best_pred = checkpoint['test_best_pred']
 
         print("=> loaded checkpoint '{}' (epoch {})"
-              .format(self.resume_path, checkpoint['epoch']))
+              .format(self.resume_path, checkpoint['epoch']))