Introducing RF computation considering models with specific input resolution. Previou...
authorA. Unique TensorFlower <gardener@tensorflow.org>
Tue, 16 Jan 2018 23:53:35 +0000 (15:53 -0800)
committerTensorFlower Gardener <gardener@tensorflow.org>
Tue, 16 Jan 2018 23:59:31 +0000 (15:59 -0800)
commit7e13a9ea3709301186b946a8c1f864e1245e6271
treec23f7516170a3e7d953257795b50b1e755d5a98a
parentccbd14b741e6efbe51769f0f1b9cb3719c42c23b
Introducing RF computation considering models with specific input resolution. Previously, the input resolution was not taken into account, which led to undefined padding for many well-known models (since those rely on SAME padding, and in some cases SAME padding depends on input resolution).

This change also redesigns many aspects of the topological sorting and layer parsing functions, introducing new modules and tests.

PiperOrigin-RevId: 182124694
tensorflow/contrib/receptive_field/BUILD
tensorflow/contrib/receptive_field/python/util/examples/rf_benchmark.py
tensorflow/contrib/receptive_field/python/util/graph_compute_order.py
tensorflow/contrib/receptive_field/python/util/graph_compute_order_test.py [new file with mode: 0644]
tensorflow/contrib/receptive_field/python/util/parse_layer_parameters.py [new file with mode: 0644]
tensorflow/contrib/receptive_field/python/util/parse_layer_parameters_test.py [new file with mode: 0644]
tensorflow/contrib/receptive_field/python/util/receptive_field.py
tensorflow/contrib/receptive_field/python/util/receptive_field_test.py