model-optimizer/mo/front/common/partial_infer/multi_box_detection_test.py

   1 """
   2  Copyright (c) 2018-2019 Intel Corporation
   3
   4  Licensed under the Apache License, Version 2.0 (the "License");
   5  you may not use this file except in compliance with the License.
   6  You may obtain a copy of the License at
   7
   8       http://www.apache.org/licenses/LICENSE-2.0
   9
  10  Unless required by applicable law or agreed to in writing, software
  11  distributed under the License is distributed on an "AS IS" BASIS,
  12  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
  13  See the License for the specific language governing permissions and
  14  limitations under the License.
  15 """
  16
  17 import unittest
  18
  19 import numpy as np
  20
  21 from mo.front.common.partial_infer.multi_box_detection import multi_box_detection_infer
  22 from mo.graph.graph import Node
  23 from mo.utils.unittest.graph import build_graph
  24
  25 nodes_attributes = {'node_1': {'value': None, 'kind': 'data'},
  26                     'node_2': {'value': None, 'kind': 'data'},
  27                     'node_3': {'value': None, 'kind': 'data'},
  28                     'detection_output_1': {'type': 'DetectionOutput', 'value': None, 'kind': 'op'},
  29                     'node_4': {'value': None, 'kind': 'data'}
  30                     }
  31
  32
  33 class TestMultiBoxDetectionInfer(unittest.TestCase):
  34     def test_prior_box_infer_ideal(self):
  35         graph = build_graph(nodes_attributes,
  36                             [('node_1', 'detection_output_1'),
  37                              ('node_2', 'detection_output_1'),
  38                              ('node_3', 'detection_output_1'),
  39                              ('detection_output_1', 'node_4')],
  40                             {'node_1': {'shape': np.array([1, 34928])},
  41                              'node_2': {'shape': np.array([1, 183372])},
  42                              'node_3': {'shape': np.array([1, 2, 34928])},
  43                              'detection_output_1': {"background_label_id": "0", "clip": "1",
  44                                                     "code_type": "caffe.PriorBoxParameter.CENTER_SIZE",
  45                                                     "confidence_threshold": "0.01", "keep_top_k": "200",
  46                                                     "nms_threshold": "0.5", "num_classes": "21",
  47                                                     "share_location": "1", "top_k": "200",
  48                                                     "variance_encoded_in_target": "0"},
  49                              'node_4': {'shape': np.array([1, 1, 200, 7])},
  50                              })
  51
  52         multi_box_detection_node = Node(graph, 'detection_output_1')
  53         print(multi_box_detection_node)
  54
  55         multi_box_detection_infer(multi_box_detection_node)
  56         exp_shape = np.array([1, 1, 200, 7])
  57         res_shape = graph.node['node_4']['shape']
  58         for i in range(0, len(exp_shape)):
  59             self.assertEqual(exp_shape[i], res_shape[i])
  60
  61         self.assertEqual(multi_box_detection_node.background_label_id, '0')
  62         self.assertEqual(multi_box_detection_node.clip, '1')
  63         self.assertEqual(multi_box_detection_node.code_type, 'caffe.PriorBoxParameter.CENTER_SIZE')
  64         self.assertEqual(multi_box_detection_node.confidence_threshold, '0.01')
  65         self.assertEqual(multi_box_detection_node.keep_top_k, '200')
  66         self.assertEqual(multi_box_detection_node.nms_threshold, '0.5')
  67         self.assertEqual(multi_box_detection_node.num_classes, 21)
  68         self.assertEqual(multi_box_detection_node.share_location, '1')
  69         self.assertEqual(multi_box_detection_node.top_k, '200')
  70         self.assertEqual(multi_box_detection_node.variance_encoded_in_target, '0')
  71
  72     def test_prior_box_infer_without_top_k(self):
  73         graph = build_graph(nodes_attributes,
  74                             [('node_1', 'detection_output_1'),
  75                              ('node_2', 'detection_output_1'),
  76                              ('node_3', 'detection_output_1'),
  77                              ('detection_output_1', 'node_4')],
  78                             {'node_1': {'shape': np.array([1, 34928])},
  79                              'node_2': {'shape': np.array([1, 183372])},
  80                              'node_3': {'shape': np.array([1, 2, 34928])},
  81                              'detection_output_1': {"background_label_id": "0", "clip": "1",
  82                                                     "code_type": "caffe.PriorBoxParameter.CENTER_SIZE",
  83                                                     "confidence_threshold": "0.01", "keep_top_k": -1,
  84                                                     "nms_threshold": "0.5", "num_classes": "21",
  85                                                     "share_location": "1", "top_k": -1,
  86                                                     "variance_encoded_in_target": "0"},
  87                              'node_4': {'shape': np.array([1, 1, 69856, 7])},
  88                              })
  89
  90         multi_box_detection_node = Node(graph, 'detection_output_1')
  91
  92         multi_box_detection_infer(multi_box_detection_node)
  93         exp_shape = np.array([1, 1, 8732, 7])
  94         res_shape = graph.node['node_4']['shape']
  95         for i in range(0, len(exp_shape)):
  96             self.assertEqual(exp_shape[i], res_shape[i])
  97
  98         self.assertEqual(multi_box_detection_node.background_label_id, '0')
  99         self.assertEqual(multi_box_detection_node.clip, '1')
 100         self.assertEqual(multi_box_detection_node.code_type, 'caffe.PriorBoxParameter.CENTER_SIZE')
 101         self.assertEqual(multi_box_detection_node.confidence_threshold, '0.01')
 102         self.assertEqual(multi_box_detection_node.keep_top_k, 8732)
 103         self.assertEqual(multi_box_detection_node.nms_threshold, '0.5')
 104         self.assertEqual(multi_box_detection_node.num_classes, 21)
 105         self.assertEqual(multi_box_detection_node.share_location, '1')
 106         self.assertEqual(multi_box_detection_node.top_k, -1)
 107         self.assertEqual(multi_box_detection_node.variance_encoded_in_target, '0')
 108
 109     def test_prior_box_infer_raise_error(self):
 110         graph = build_graph(nodes_attributes,
 111                             [('node_1', 'detection_output_1'),
 112                              ('node_2', 'detection_output_1'),
 113                              ('node_3', 'detection_output_1'),
 114                              ('detection_output_1', 'node_4')],
 115                             {'node_1': {'shape': np.array([1, 34928])},
 116                              'node_2': {'shape': np.array([1, 183372])},
 117                              'node_3': {'shape': np.array([1, 3, 34928])},
 118                              'detection_output_1': {"background_label_id": "0", "clip": "1",
 119                                                     "code_type": "caffe.PriorBoxParameter.CENTER_SIZE",
 120                                                     "confidence_threshold": "0.01", "keep_top_k": -1,
 121                                                     "nms_threshold": "0.5", "num_classes": "21",
 122                                                     "share_location": "1", "top_k": -1,
 123                                                     "variance_encoded_in_target": 0},
 124                              'node_4': {'shape': np.array([1, 1, 69856, 7])},
 125                              })
 126
 127         multi_box_detection_node = Node(graph, 'detection_output_1')
 128
 129         self.assertIsNone(multi_box_detection_infer(multi_box_detection_node))