0.13.2 (#1102)

* Increment version to 0.13.2 (#1095)

* Bump version

* Update configs

* dapricot test set (#1096)

* cherry-picked dapricot test commits from 1088

* correct checksum filename

* Coco (#1097)

* cherry-picking commits from 1085, excluding the commit merging in dev branch

* adding coco tests, skipping if not available locally

* adding note to docs about apricot class indexing

* updated checksum after new upload to s3

Co-authored-by: ng390 <neal.gupta@twosixlabs.com>

armory/__init__.py

@@ -8,7 +8,7 @@
 
 
 # Semantic Version
-__version__ = "0.13.1"
+__version__ = "0.13.2"
 
 
 # Submodule imports

armory/baseline_models/tf_graph/mscoco_frcnn.py

@@ -9,63 +9,29 @@
 import tensorflow as tf
 
 
-class TensorFlowFasterRCNNOneIndexed(TensorFlowFasterRCNN):
-    """
-    This is an MSCOCO pre-trained model. Note that the inherited TensorFlowFasterRCMM class
-    outputs 0-indexed classes, while this wrapper class outputs 1-indexed classes. A label map can be found at
-    https://github.com/tensorflow/models/blob/master/research/object_detection/data/mscoco_label_map.pbtxt
-
-    This model only performs inference and is not trainable. To train
-    or fine-tune this model, please follow instructions at
-    https://github.com/tensorflow/models/blob/master/research/object_detection/g3doc/tf1.md
-    """
-
-    def __init__(self, images):
-        super().__init__(
-            images,
-            model=None,
-            filename="faster_rcnn_resnet50_coco_2018_01_28",
-            url="http://download.tensorflow.org/models/object_detection/faster_rcnn_resnet50_coco_2018_01_28.tar.gz",
-            sess=None,
-            is_training=False,
-            clip_values=(0, 1),
-            channels_first=False,
-            preprocessing_defences=None,
-            postprocessing_defences=None,
-            attack_losses=(
-                "Loss/RPNLoss/localization_loss",
-                "Loss/RPNLoss/objectness_loss",
-                "Loss/BoxClassifierLoss/localization_loss",
-                "Loss/BoxClassifierLoss/classification_loss",
-            ),
-        )
-
-    def compute_loss(self, x, y):
-        raise NotImplementedError
-
-    def loss_gradient(self, x, y, **kwargs):
-        y_zero_indexed = []
-        for y_dict in y:
-            y_dict_zero_indexed = y_dict.copy()
-            y_dict_zero_indexed["labels"] = y_dict_zero_indexed["labels"] - 1
-            y_zero_indexed.append(y_dict_zero_indexed)
-        return super().loss_gradient(x, y_zero_indexed, **kwargs)
-
-    def predict(self, x, **kwargs):
-        list_of_zero_indexed_pred_dicts = super().predict(x, **kwargs)
-        list_of_one_indexed_pred_dicts = []
-        for img_pred_dict in list_of_zero_indexed_pred_dicts:
-            zero_indexed_pred_labels = img_pred_dict["labels"]
-            img_pred_dict["labels"] = zero_indexed_pred_labels + 1
-            list_of_one_indexed_pred_dicts.append(img_pred_dict)
-        return list_of_one_indexed_pred_dicts
-
-
 def get_art_model(model_kwargs, wrapper_kwargs, weights_file=None):
     # APRICOT inputs should have shape (1, None, None, 3) while DAPRICOT inputs have shape
     # (3, None, None, 3)
     images = tf.placeholder(
         tf.float32, shape=(model_kwargs.get("batch_size", 1), None, None, 3)
     )
-    model = TensorFlowFasterRCNNOneIndexed(images)
+    model = TensorFlowFasterRCNN(
+        images,
+        model=None,
+        filename="faster_rcnn_resnet50_coco_2018_01_28",
+        url="http://download.tensorflow.org/models/object_detection/faster_rcnn_resnet50_coco_2018_01_28.tar.gz",
+        sess=None,
+        is_training=False,
+        clip_values=(0, 1),
+        channels_first=False,
+        preprocessing_defences=None,
+        postprocessing_defences=None,
+        attack_losses=(
+            "Loss/RPNLoss/localization_loss",
+            "Loss/RPNLoss/objectness_loss",
+            "Loss/BoxClassifierLoss/localization_loss",
+            "Loss/BoxClassifierLoss/classification_loss",
+        ),
+    )
+
     return model