Add RESNET_ARCHITECTURE variable and change config.py to its original…

… state

README.md

@@ -8,6 +8,7 @@ I did almost no changes to the original code, except for:
 * `VALIDATION_STEPS` and `STEPS_PER_EPOCH` are now forced to depend on the dataset size, hardcoded.
 * `multiprocessing=False`, hardcoded.
 * [@John1231983]'s changed from [this PR](https://github.com/killthekitten/kaggle-ds-bowl-2018-baseline/pull/1).
+* Added `RESNET_ARCHITECTURE` variable to the config (`resnet50` or `resnet101` while 101 comes with a default config).
 
 ## Quick Start
 
@@ -42,7 +43,7 @@ CUDA_VISIBLE_DEVICES="0" python inference.py
 
 This will create `submission.csv` in the repo and overwrite the old one (you're welcome to fix this with a PR).
 
-6. Submit! You should get around 0.342 score on LB after 100 epochs. 
+6. Submit! You should get around 0.4 score on LB after 100 epochs.
 
 ## What's else inside?
 

bowl_config.py

@@ -32,8 +32,15 @@ class BowlConfig(Config):
 
     # use small validation steps since the epoch is small
     VALIDATION_STEPS = 5
-    
+
     USE_MINI_MASK = True
-
+
+    MAX_GT_INSTANCES = 256
+
+    DETECTION_MAX_INSTANCES = 512
+
+    RESNET_ARCHITECTURE = "resnet50"
+
+
 bowl_config = BowlConfig()
 bowl_config.display()

config.py

@@ -41,7 +41,7 @@ class Config(object):
     # Validation stats are also calculated at each epoch end and they
     # might take a while, so don't set this too small to avoid spending
     # a lot of time on validation stats.
-    STEPS_PER_EPOCH = 670
+    STEPS_PER_EPOCH = 1000
 
     # Number of validation steps to run at the end of every training epoch.
     # A bigger number improves accuracy of validation stats, but slows
@@ -87,8 +87,8 @@ class Config(object):
     # Images are resized such that the smallest side is >= IMAGE_MIN_DIM and
     # the longest side is <= IMAGE_MAX_DIM. In case both conditions can't
     # be satisfied together the IMAGE_MAX_DIM is enforced.
-    IMAGE_MIN_DIM = 512
-    IMAGE_MAX_DIM = 512
+    IMAGE_MIN_DIM = 800
+    IMAGE_MAX_DIM = 1024
     # If True, pad images with zeros such that they're (max_dim by max_dim)
     IMAGE_PADDING = True  # currently, the False option is not supported
 
@@ -111,14 +111,14 @@ class Config(object):
     MASK_SHAPE = [28, 28]
 
     # Maximum number of ground truth instances to use in one image
-    MAX_GT_INSTANCES = 256
+    MAX_GT_INSTANCES = 100
 
     # Bounding box refinement standard deviation for RPN and final detections.
     RPN_BBOX_STD_DEV = np.array([0.1, 0.1, 0.2, 0.2])
     BBOX_STD_DEV = np.array([0.1, 0.1, 0.2, 0.2])
 
     # Max number of final detections
-    DETECTION_MAX_INSTANCES = 512
+    DETECTION_MAX_INSTANCES = 100
 
     # Minimum probability value to accept a detected instance
     # ROIs below this threshold are skipped
@@ -144,6 +144,8 @@ class Config(object):
     # train the RPN.
     USE_RPN_ROIS = True
 
+    RESNET_ARCHITECTURE = "resnet101"
+
     def __init__(self):
         """Set values of computed attributes."""
         # Effective batch size

model.py

@@ -1816,7 +1816,7 @@ def build(self, mode, config):
         # Bottom-up Layers
         # Returns a list of the last layers of each stage, 5 in total.
         # Don't create the thead (stage 5), so we pick the 4th item in the list.
-        _, C2, C3, C4, C5 = resnet_graph(input_image, "resnet50", stage5=True)
+        _, C2, C3, C4, C5 = resnet_graph(input_image, config.RESNET_ARCHITECTURE, stage5=True)
         # Top-down Layers
         # TODO: add assert to varify feature map sizes match what's in config
         P5 = KL.Conv2D(256, (1, 1), name='fpn_c5p5')(C5)
@@ -2224,10 +2224,10 @@ def train(self, train_dataset, val_dataset, learning_rate, epochs, layers):
             workers = 0
         else:
             workers = max(self.config.BATCH_SIZE // 2, 2)
-        
+
         steps_per_epoch = np.ceil(len(train_dataset.image_ids) / self.config.BATCH_SIZE)
         validation_steps = np.ceil(len(train_dataset.image_ids) / self.config.BATCH_SIZE)
-        
+
         self.keras_model.fit_generator(
             train_generator,
             initial_epoch=self.epoch,