0.0.2.1 debug

Author: akiozihao

commandline.txt

@@ -1,2 +1,5 @@
 DATASET_ROOT_PATH = "/home/akio/PycharmProjects/rat-detect/data"
-
+TRAIN_IMAGES_PATH="/home/akio/PycharmProjects/rat-detect/data/train_images.txt"
+TRAIN_XMLS_PATH = "/home/akio/PycharmProjects/rat-detect/data/train_xmls.txt"
+VAL_IMAGES_PATH = "/home/akio/PycharmProjects/rat-detect/data/val_images.txt"
+VAL_XMLS_PATH = "/home/akio/PycharmProjects/rat-detect/data/val_xmls.txt"

config.py

@@ -11,17 +11,10 @@ def main(root):
         sys.exit(1)
 
     images = glob.glob(f"{root}/*/*.jpg")
-    xmls = glob.glob(f"{root}/*/*.xml")
-
-    def map_f(x):
-        return root + x[1:]
-
-    map(map_f, images)
-    map(map_f, xmls)
 
+    xmls = [image[:-3] + 'xml' for image in images]
     images = np.array(images)
     xmls = np.array(xmls)
-    assert len(images) == len(xmls)
     total_size = len(images)
 
     train_size = int(0.7 * total_size)

models/ssd/datasets/generate_data_path.py

@@ -9,26 +9,29 @@
 
 class RATDataset(Dataset):
     def __init__(self,
-                 root,
                  images_path,
                  xmls_path,
-
                  transform=None,
                  target_transform=None,
                  is_test=False):
-        self.root = root
         self.transform = transform
         self.target_transform = target_transform
 
         self.images = RATDataset._read_images_path(images_path)
         self.xmls = RATDataset._read_xmls_path(xmls_path)
+        assert len(self.images) == len(self.xmls)
         self.class_names = ('BACKGROUND', 'rat')
 
         self.class_dict = {class_name: i for i, class_name in enumerate(self.class_names)}
+        self.num = 0
 
     def __getitem__(self, index):
+        self.num += 1
+
         image_path = self.images[index]
+
         xml_path = self.xmls[index]
+        print('image_path:', image_path,'||xml_path:',xml_path)
         boxes, labels = self._get_annotation(xml_path)
         image = self._read_image(image_path)
         if self.transform:
@@ -38,18 +41,18 @@ def __getitem__(self, index):
         return image, boxes, labels
 
     def __len__(self):
-        return len(self.ids)
+        return len(self.images)
 
     def get_image(self, index):
-        image_id = self.ids[index]
-        image = self._read_image(image_id)
+        image_path = self.images[index]
+        image = self._read_image(image_path)
         if self.transform:
             image, _ = self.transform(image)
         return image
 
     def get_annotation(self, index):
-        image_id = self.ids[index]
-        return image_id, self._get_annotation(image_id)
+        image_path = self.images[index]
+        return image_path, self._get_annotation(image_path)
 
     def _get_annotation(self, xml_path):
         objects = ET.parse(xml_path).findall("object")
@@ -82,12 +85,12 @@ def _read_images_path(images_path):
         with open(images_path) as f:
             for line in f:
                 images.append(line.rstrip())
-        return images[:-1]
+        return images
 
     @staticmethod
     def _read_xmls_path(xmls_path):
         xmls = []
         with open(xmls_path) as f:
             for line in f:
                 xmls.append(line.rstrip())
-        return xmls[:-1]
+        return xmls

models/ssd/datasets/rat_dataset.py

@@ -10,6 +10,7 @@
 
 
 def intersect(box_a, box_b):
+    # print('box_a',box_a,'||box_b',box_b)
     max_xy = np.minimum(box_a[:, 2:], box_b[2:])
     min_xy = np.maximum(box_a[:, :2], box_b[:2])
     inter = np.clip((max_xy - min_xy), a_min=0, a_max=np.inf)

models/transforms/transforms.py

@@ -17,13 +17,17 @@
 from models.ssd.config import mobilenetv1_ssd_config
 from models.ssd.data_preprocessing import TrainAugmentation, TestTransform
 
+import config
+
 parser = argparse.ArgumentParser(
     description='Single Shot MultiBox Detector Training With Pytorch')
 
-parser.add_argument('--datasets', nargs='+', help='Dataset directory path')
-parser.add_argument('--validation_dataset', help='Dataset directory path')
+parser.add_argument('--train_images_path', default=config.TRAIN_IMAGES_PATH, help='train_images_path')
+parser.add_argument('--train_xmls_path', default=config.TRAIN_XMLS_PATH, help='train_xmls_path')
+parser.add_argument('--val_images_path', default=config.VAL_IMAGES_PATH, help='val_images_path')
+parser.add_argument('--val_xmls_path', default=config.VAL_XMLS_PATH, help='val_xmls_path')
 
-parser.add_argument('--net', default='mb1-ssd',
+parser.add_argument('--net', default='mb1-ssd-lite',
                     help='The network architecture')
 parser.add_argument('--freeze_base_net', action='store_true',
                     help='Freeze base net layers.')
@@ -64,11 +68,11 @@
                     help='T_max value for Cosine Annealing Scheduler.')
 
 # Train params
-parser.add_argument('--batch_size', default=32, type=int,
+parser.add_argument('--batch_size', default=8, type=int,
                     help='Batch size for training')
 parser.add_argument('--num_epochs', default=120, type=int,
                     help='the number epochs')
-parser.add_argument('--num_workers', default=4, type=int,
+parser.add_argument('--num_workers', default=1, type=int,
                     help='Number of workers used in dataloading')
 parser.add_argument('--validation_epochs', default=5, type=int,
                     help='the number epochs')
@@ -96,6 +100,7 @@ def train(loader, net, criterion, optimizer, device, debug_steps=100, epoch=-1):
     running_regression_loss = 0.0
     running_classification_loss = 0.0
     for i, data in enumerate(loader):
+        print(f"i:{i}")
         images, boxes, labels = data
         images = images.to(device)
         boxes = boxes.to(device)
@@ -168,20 +173,23 @@ def test(loader, net, criterion, device):
     test_transform = TestTransform(config.image_size, config.image_mean, config.image_std)
 
     logging.info("Prepare training datasets.")
-    datasets = []
-
-    for dataset_path in args.datasets:
-        dataset = RATDataset(dataset_path, transform=train_transform, target_transform=target_transform)
-        num_classes = len(dataset.class_names)
-        datasets.append(dataset)
-    train_datasets = ConcatDataset(datasets)
+    # datasets = []
+
+    # for dataset_path in args.datasets:
+    #     dataset = RATDataset(dataset_path, transform=train_transform, target_transform=target_transform)
+    #     num_classes = len(dataset.class_names)
+    #     datasets.append(dataset)
+    # train_datasets = ConcatDataset(datasets)
+    train_datasets = RATDataset(args.train_images_path, args.train_xmls_path, transform=train_transform,
+                                target_transform=target_transform)
+    num_classes = len(train_datasets.class_names)
     logging.info(f"Train dataset size :{len(train_datasets)}")
     logging.info(train_datasets)
     train_loader = DataLoader(train_datasets, args.batch_size,
-                                  num_workers=args.num_workers, shuffle=True)
+                              num_workers=args.num_workers, shuffle=True)
 
     logging.info("Prepare Validation datasets.")
-    val_dataset = RATDataset(args.validation_dataset, transform=test_transform,
+    val_dataset = RATDataset(args.val_images_path, args.val_xmls_path, transform=test_transform,
                              target_transform=target_transform, is_test=True)
     logging.info(val_dataset)
     val_loader = DataLoader(val_dataset, args.batch_size,
@@ -279,4 +287,3 @@ def test(loader, net, criterion, device):
             model_path = os.path.join(args.checkpoint_folder, f"{args.net}-Epoch-{epoch}-Loss-{val_loss}.pth")
             net.save(model_path)
             logging.info(f"Saved model {model_path}")
-