115 lines
4.1 KiB
Python
115 lines
4.1 KiB
Python
# Copyright (C) 2018-2021 Intel Corporation
|
|
# SPDX-License-Identifier: Apache-2.0
|
|
|
|
import numpy as np
|
|
|
|
from mo.front.common.layout import get_width_dim, get_height_dim
|
|
from mo.front.extractor import attr_getter, bool_to_str
|
|
from mo.graph.graph import Node, Graph
|
|
from mo.ops.op import Op
|
|
|
|
|
|
class PriorBoxOp(Op):
|
|
op = 'PriorBox'
|
|
|
|
def __init__(self, graph: Graph, attrs: dict):
|
|
mandatory_props = {
|
|
'type': self.op,
|
|
'op': self.op,
|
|
'version': 'opset1',
|
|
'flip': True,
|
|
'clip': True,
|
|
'scale_all_sizes': True,
|
|
'max_size': np.array([]),
|
|
'min_size': np.array([]),
|
|
'aspect_ratio': np.array([]),
|
|
'density': np.array([]),
|
|
'fixed_size': np.array([]),
|
|
'fixed_ratio': np.array([]),
|
|
'in_ports_count': 2,
|
|
'out_ports_count': 1,
|
|
'type_infer': self.type_infer,
|
|
'infer': self.priorbox_infer
|
|
}
|
|
super().__init__(graph, mandatory_props, attrs)
|
|
|
|
def supported_attrs(self):
|
|
return [
|
|
'min_size',
|
|
'max_size',
|
|
'aspect_ratio',
|
|
'flip',
|
|
'clip',
|
|
'variance',
|
|
'img_size',
|
|
'img_h',
|
|
'img_w',
|
|
'step',
|
|
'step_h',
|
|
'step_w',
|
|
'offset',
|
|
'density',
|
|
'fixed_size',
|
|
'fixed_ratio',
|
|
]
|
|
|
|
def backend_attrs(self):
|
|
return [
|
|
('flip', lambda node: int(node.flip)), # We need to convert this boolean attribute value to int to keep
|
|
# forward compatibility with IE 2021.2
|
|
('clip', lambda node: int(node.clip)), # We need to convert this boolean attribute value to int to keep
|
|
# forward compatibility with IE 2021.2
|
|
'step',
|
|
'offset',
|
|
('scale_all_sizes', lambda node: bool_to_str(node, 'scale_all_sizes')),
|
|
('min_size', lambda node: attr_getter(node, 'min_size')),
|
|
('max_size', lambda node: attr_getter(node, 'max_size')),
|
|
('aspect_ratio', lambda node: attr_getter(node, 'aspect_ratio')),
|
|
('variance', lambda node: attr_getter(node, 'variance')),
|
|
('density', lambda node: attr_getter(node, 'density')),
|
|
('fixed_size', lambda node: attr_getter(node, 'fixed_size')),
|
|
('fixed_ratio', lambda node: attr_getter(node, 'fixed_ratio')),
|
|
]
|
|
|
|
@staticmethod
|
|
def type_infer(node):
|
|
node.out_port(0).set_data_type(np.float32)
|
|
|
|
@staticmethod
|
|
def priorbox_infer(node: Node):
|
|
layout = node.graph.graph['layout']
|
|
data_shape = node.in_node(0).shape
|
|
|
|
# calculate all different aspect_ratios (the first one is always 1)
|
|
# in aspect_ratio 1/x values will be added for all except 1 if flip is True
|
|
ar_seen = [1.0]
|
|
ar_seen.extend(node.aspect_ratio.copy())
|
|
if node.flip:
|
|
for s in node.aspect_ratio:
|
|
ar_seen.append(1.0 / s)
|
|
|
|
ar_seen = np.unique(np.array(ar_seen).round(decimals=6))
|
|
|
|
num_ratios = 0
|
|
if len(node.min_size) > 0:
|
|
num_ratios = len(ar_seen) * len(node.min_size)
|
|
|
|
if node.has_valid('fixed_size') and len(node.fixed_size) > 0:
|
|
num_ratios = len(ar_seen) * len(node.fixed_size)
|
|
|
|
if node.has_valid('density') and len(node.density) > 0:
|
|
for d in node.density:
|
|
if node.has_valid('fixed_ratio') and len(node.fixed_ratio) > 0:
|
|
num_ratios = num_ratios + len(node.fixed_ratio) * (pow(d, 2) - 1)
|
|
else:
|
|
num_ratios = num_ratios + len(ar_seen) * (pow(d, 2) - 1)
|
|
|
|
num_ratios = num_ratios + len(node.max_size)
|
|
|
|
if node.has_and_set('V10_infer'):
|
|
assert node.in_node(0).value is not None
|
|
node.out_node(0).shape = np.array([2, np.prod(node.in_node(0).value) * num_ratios * 4], dtype=np.int64)
|
|
else:
|
|
res_prod = data_shape[get_height_dim(layout, 4)] * data_shape[get_width_dim(layout, 4)] * num_ratios * 4
|
|
node.out_node(0).shape = np.array([1, 2, res_prod], dtype=np.int64)
|