1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
|
#
# Copyright (C) 2019 The Android Open Source Project
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
#
layout = BoolScalar("layout", False) # NHWC
# Operation 1, GENERATE_PROPOSALS
scores = Input("scores", "TENSOR_FLOAT32", "{1, 1, 1, 1}")
deltas = Input("deltas", "TENSOR_FLOAT32", "{1, 1, 1, 4}")
anchors = Input("anchors", "TENSOR_FLOAT32", "{1, 4}")
image = Input("imageInfo", "TENSOR_FLOAT32", "{1, 2}")
scoresOut_1 = Output("scores", "TENSOR_FLOAT32", "{0}")
roiOut_1 = Internal("roi", "TENSOR_FLOAT32", "{0, 4}")
batchOut_1 = Internal("batches", "TENSOR_INT32", "{0}")
model = Model("zero_sized").Operation("GENERATE_PROPOSALS", scores, deltas, anchors, image, 1.0, 1.0, -1, -1, 0.3, 10.0, layout).To(scoresOut_1, roiOut_1, batchOut_1)
# Operation 2, ROI_ALIGN
feature = Input("featureMap", "TENSOR_FLOAT32", "{1, 1, 1, 1}")
featureOut_2 = Internal("scores", "TENSOR_FLOAT32", "{0, 2, 2, 1}")
model = model.Operation("ROI_ALIGN", feature, roiOut_1, batchOut_1, 2, 2, 1.0, 1.0, 4, 4, layout).To(featureOut_2)
# Operation 3, FULLY_CONNECTED
weights_3 = Parameter("weights", "TENSOR_FLOAT32", "{8, 4}", [1] * 32)
bias_3 = Parameter("bias", "TENSOR_FLOAT32", "{8}", [1] * 8)
deltaOut_3 = Internal("delta", "TENSOR_FLOAT32", "{0, 8}")
model = model.Operation("FULLY_CONNECTED", featureOut_2, weights_3, bias_3, 0).To(deltaOut_3)
# Operation 4, FULLY_CONNECTED
weights_4 = Parameter("weights", "TENSOR_FLOAT32", "{2, 4}", [1] * 8)
bias_4 = Parameter("bias", "TENSOR_FLOAT32", "{2}", [1] * 2)
scoresOut_4 = Internal("scores", "TENSOR_FLOAT32", "{0, 2}")
model = model.Operation("FULLY_CONNECTED", featureOut_2, weights_4, bias_4, 0).To(scoresOut_4)
# Operation 5, AXIS_ALIGNED_BBOX_TRANSFORM
roiOut_5 = Internal("roi", "TENSOR_FLOAT32", "{0, 8}")
model = model.Operation("AXIS_ALIGNED_BBOX_TRANSFORM", roiOut_1, deltaOut_3, batchOut_1, image).To(roiOut_5)
# Operation 6, BOX_WITH_NMS_LIMIT
scoresOut_6 = Output("scores", "TENSOR_FLOAT32", "{0}")
roiOut_6 = Output("roi", "TENSOR_FLOAT32", "{0, 4}")
classOut_6 = Output("classes", "TENSOR_INT32", "{0}")
batchOut_6 = Output("batches", "TENSOR_INT32", "{0}")
model = model.Operation("BOX_WITH_NMS_LIMIT", scoresOut_4, roiOut_5, batchOut_1, 0.1, -1, 0, 0.3, 1.0, 0.1).To(scoresOut_6, roiOut_6, classOut_6, batchOut_6)
quant8 = DataTypeConverter().Identify({
scores: ("TENSOR_QUANT8_ASYMM", 0.1, 128),
deltas: ("TENSOR_QUANT8_ASYMM", 0.1, 128),
anchors: ("TENSOR_QUANT16_SYMM", 0.125, 0),
image: ("TENSOR_QUANT16_ASYMM", 0.125, 0),
scoresOut_1: ("TENSOR_QUANT8_ASYMM", 0.1, 128),
roiOut_1: ("TENSOR_QUANT16_ASYMM", 0.125, 0),
feature: ("TENSOR_QUANT8_ASYMM", 0.1, 128),
featureOut_2: ("TENSOR_QUANT8_ASYMM", 0.1, 128),
weights_3: ("TENSOR_QUANT8_ASYMM", 0.1, 128),
bias_3: ("TENSOR_INT32", 0.01, 0),
deltaOut_3: ("TENSOR_QUANT8_ASYMM", 0.1, 128),
weights_4: ("TENSOR_QUANT8_ASYMM", 0.1, 128),
bias_4: ("TENSOR_INT32", 0.01, 0),
scoresOut_4: ("TENSOR_QUANT8_ASYMM", 0.1, 128),
roiOut_5: ("TENSOR_QUANT16_ASYMM", 0.125, 0),
scoresOut_6: ("TENSOR_QUANT8_ASYMM", 0.1, 128),
roiOut_6: ("TENSOR_QUANT16_ASYMM", 0.125, 0),
})
Example({
# Inputs that will lead to zero-sized output of GENERATE_PROPOSALS
scores: [0.5],
deltas: [0, 0, -10, -10],
anchors: [0, 0, 10, 10],
image: [32, 32],
feature: [1],
# Dummy outputs
scoresOut_1: [0],
scoresOut_6: [0],
roiOut_6: [0],
classOut_6: [0],
batchOut_6: [0],
}).AddVariations("relaxed", "float16", quant8)
|
