Test for training and inference for dense_vnet_abdominal_ct

.gitlab-ci.yml

@@ -29,6 +29,7 @@ testjob:
     - dev
     - tags
     - 195-add-evaluation-action-3
+    - 192-model_zoo_tests
   script:
     # !!kill coverage in case of hanging processes
     - if pgrep coverage; then pkill -f coverage; fi
@@ -200,6 +201,7 @@ quicktest:
     - 221-add-changelog-entry-for-version-0.2.2
     - 223-put-bug-fixes-under-fixed-header-in-changelog
     - 195-add-evaluation-action-3
+    - 192-model_zoo_tests
   script:
     # print system info
     - which nvidia-smi

niftynet/application/base_application.py

@@ -18,16 +18,22 @@
 EVAL = "evaluation"
 
 
-class SingletonApplication(type):
-    _instances = None
 
+application_singleton_instance = None # global so it can be reset
+class SingletonApplication(type):
     def __call__(cls, *args, **kwargs):
-        if cls._instances is None:
-            cls._instances = \
+        global application_singleton_instance
+        if application_singleton_instance is None:
+            application_singleton_instance = \
                 super(SingletonApplication, cls).__call__(*args, **kwargs)
         # else:
         #     raise RuntimeError('application instance already started.')
-        return cls._instances
+        return application_singleton_instance
+
+    @classmethod
+    def clear(cls):
+        global application_singleton_instance
+        application_singleton_instance = None
 
 
 class BaseApplication(with_metaclass(SingletonApplication, object)):

niftynet/layer/resampler.py

@@ -466,10 +466,12 @@ def _binary_neighbour_ids(spatial_rank):
             for i in range(2 ** spatial_rank)]
 
 
-@tf.RegisterGradient('FloorMod')
-def _floormod_grad(op, grad):
-    return [None, None]
-
+try: # Some tf versions have this defined already
+    @tf.RegisterGradient('FloorMod')
+    def _floormod_grad(op, grad):
+        return [None, None]
+except:
+    pass
 
 SUPPORTED_INTERPOLATION = {'BSPLINE', 'LINEAR', 'NEAREST', 'IDW'}
 

tests/model_zoo_test.py

@@ -0,0 +1,60 @@
+# -*- coding: utf-8 -*-
+from __future__ import absolute_import
+from __future__ import division
+from __future__ import print_function
+
+import os, sys, unittest
+
+import tensorflow as tf
+
+from niftynet.utilities.download import download
+from niftynet.utilities.niftynet_global_config import NiftyNetGlobalConfig
+from niftynet import main as niftynet_main
+from niftynet.application.base_application import SingletonApplication
+
+MODEL_HOME = NiftyNetGlobalConfig().get_niftynet_home_folder()
+
+def net_run_with_sys_argv(argv):
+    cache = sys.argv
+    sys.argv = argv
+    niftynet_main()
+    sys.argv = cache
+
+class ModelZooTestMixin(object):
+    def test_train(self):
+        self.assertTrue(download('dense_vnet_abdominal_ct_model_zoo', True))
+        SingletonApplication.clear()
+        net_run_with_sys_argv(['net_run', '-a', self.application, '-c', 
+            os.path.join(MODEL_HOME, 'extensions', self.location, self.config), 
+            'train', '--max_iter', '2'])
+        checkpoint = os.path.join(MODEL_HOME, 'models', self.location, 'models', 'model.ckpt-1.index')
+        self.assertTrue(os.path.exists(checkpoint))
+        self.check_train()    
+
+    def test_inference(self):
+        self.assertTrue(download('dense_vnet_abdominal_ct_model_zoo', True))
+        SingletonApplication.clear()
+        net_run_with_sys_argv(['net_run', '-a', self.application, '-c', 
+            os.path.join(MODEL_HOME, 'extensions', self.location, self.config), 
+            'inference'])
+        output = os.path.join(MODEL_HOME, 'models', self.location, self.expected_output)
+        self.assertTrue(os.path.exists(output))
+        self.check_inference()    
+
+    def check_inference(self):
+        pass
+
+    def check_train(self):
+        pass
+
+
+@unittest.skipIf(os.environ.get('QUICKTEST', "").lower() == "true", 'Skipping slow tests')
+class DenseVNetAbdominalCTModelZooTest(tf.test.TestCase, ModelZooTestMixin):
+    id = 'dense_vnet_abdominal_ct_model_zoo'
+    location = 'dense_vnet_abdominal_ct'
+    config = 'config.ini'
+    application = 'net_segment'
+    expected_output = os.path.join('segmentation_output','100__niftynet_out.nii.gz')
+
+if __name__ == "__main__":
+    tf.test.main()