-
Notifications
You must be signed in to change notification settings - Fork 2.7k
/
roialign.cc
330 lines (285 loc) · 13 KB
/
roialign.cc
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
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
/**
* Copyright (c) 2016-present, Facebook, Inc.
*
* 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.
*/
/* Modifications Copyright (c) Microsoft. */
#include "roialign.h"
#include <cmath>
#include <core/common/safeint.h>
#include "core/util/math_cpuonly.h"
#include "core/common/common.h"
#include "core/framework/tensor.h"
#include "core/platform/threadpool.h"
using namespace onnxruntime::concurrency;
namespace onnxruntime {
#define ADD_VERSIONED_TYPED_ROIALIGN_OP(data_type) \
ONNX_CPU_OPERATOR_VERSIONED_TYPED_KERNEL( \
RoiAlign, \
10, \
15, \
data_type, \
KernelDefBuilder() \
.TypeConstraint("T1", DataTypeImpl::GetTensorType<data_type>()) \
.TypeConstraint("T2", DataTypeImpl::GetTensorType<int64_t>()), \
RoiAlign<data_type>);
ADD_VERSIONED_TYPED_ROIALIGN_OP(float);
ADD_VERSIONED_TYPED_ROIALIGN_OP(double);
#define ADD_TYPED_ROIALIGN_OP(data_type) \
ONNX_CPU_OPERATOR_TYPED_KERNEL(RoiAlign, 16, data_type, \
KernelDefBuilder() \
.TypeConstraint("T1", DataTypeImpl::GetTensorType<data_type>()) \
.TypeConstraint("T2", DataTypeImpl::GetTensorType<int64_t>()), \
RoiAlign<data_type>);
ADD_TYPED_ROIALIGN_OP(float);
ADD_TYPED_ROIALIGN_OP(double);
namespace {
template <typename T>
struct PreCalc {
int64_t pos1;
int64_t pos2;
int64_t pos3;
int64_t pos4;
T w1;
T w2;
T w3;
T w4;
};
// TODO: fix the warnings
#if defined(_MSC_VER) && !defined(__clang__)
// Chance of arithmetic overflow could be reduced
#pragma warning(disable : 26451)
#endif
template <typename T>
static void PreCalcForBilinearInterpolate(const int64_t height, const int64_t width, const int64_t pooled_height,
const int64_t pooled_width, const int64_t iy_upper, const int64_t ix_upper,
T roi_start_h, T roi_start_w, T bin_size_h, T bin_size_w, int64_t roi_bin_grid_h,
int64_t roi_bin_grid_w, std::vector<PreCalc<T>>& pre_calc) {
int64_t pre_calc_index = 0;
for (int64_t ph = 0; ph < pooled_height; ph++) {
for (int64_t pw = 0; pw < pooled_width; pw++) {
for (int64_t iy = 0; iy < iy_upper; iy++) {
const T yy = roi_start_h + ph * bin_size_h +
static_cast<T>(iy + .5f) * bin_size_h / static_cast<T>(roi_bin_grid_h); // e.g., 0.5, 1.5
for (int64_t ix = 0; ix < ix_upper; ix++) {
const T xx =
roi_start_w + pw * bin_size_w + static_cast<T>(ix + .5f) * bin_size_w / static_cast<T>(roi_bin_grid_w);
T x = xx;
T y = yy;
// deal with: inverse elements are out of feature map boundary
if (y < -1.0 || y > height || x < -1.0 || x > width) {
auto& pc = pre_calc[onnxruntime::narrow<size_t>(pre_calc_index)];
pc.pos1 = 0;
pc.pos2 = 0;
pc.pos3 = 0;
pc.pos4 = 0;
pc.w1 = 0;
pc.w2 = 0;
pc.w3 = 0;
pc.w4 = 0;
pre_calc_index += 1;
continue;
}
if (y <= 0) {
y = 0;
}
if (x <= 0) {
x = 0;
}
auto y_low = static_cast<int64_t>(y);
auto x_low = static_cast<int64_t>(x);
int64_t y_high;
int64_t x_high;
if (y_low >= height - 1) {
y_high = y_low = height - 1;
y = (T)y_low;
} else {
y_high = y_low + 1;
}
if (x_low >= width - 1) {
x_high = x_low = width - 1;
x = (T)x_low;
} else {
x_high = x_low + 1;
}
T ly = y - y_low;
T lx = x - x_low;
T hy = static_cast<T>(1.) - ly;
T hx = static_cast<T>(1.) - lx;
T w1 = hy * hx;
T w2 = hy * lx;
T w3 = ly * hx;
T w4 = ly * lx;
// save weights and indeces
PreCalc<T> pc;
pc.pos1 = y_low * width + x_low;
pc.pos2 = y_low * width + x_high;
pc.pos3 = y_high * width + x_low;
pc.pos4 = y_high * width + x_high;
pc.w1 = w1;
pc.w2 = w2;
pc.w3 = w3;
pc.w4 = w4;
pre_calc[onnxruntime::narrow<size_t>(pre_calc_index)] = pc;
pre_calc_index += 1;
}
}
}
}
}
template <typename T>
void RoiAlignForward(const TensorShape& output_shape, const T* bottom_data, float spatial_scale, int64_t height,
int64_t width, int64_t sampling_ratio, const T* bottom_rois, int64_t num_roi_cols, T* top_data,
RoiAlignMode mode, bool half_pixel, const int64_t* batch_indices_ptr, ThreadPool* ttp) {
int64_t n_rois = output_shape[0];
int64_t channels = output_shape[1];
int64_t pooled_height = output_shape[2];
int64_t pooled_width = output_shape[3];
// 100 is a random chosed value, need be tuned
double cost = static_cast<double>(channels * pooled_width * pooled_height * 100);
ThreadPool::TryParallelFor(ttp, static_cast<ptrdiff_t>(n_rois), cost, [&](ptrdiff_t n, ptrdiff_t end) {
for (; n != end; ++n) {
int64_t index_n = n * channels * pooled_width * pooled_height;
const T* offset_bottom_rois = bottom_rois + n * num_roi_cols;
const auto roi_batch_ind = batch_indices_ptr[n];
// Do not using rounding; this implementation detail is critical
T offset = half_pixel ? (T)0.5 : (T)0.0;
T roi_start_w = offset_bottom_rois[0] * spatial_scale - offset;
T roi_start_h = offset_bottom_rois[1] * spatial_scale - offset;
T roi_end_w = offset_bottom_rois[2] * spatial_scale - offset;
T roi_end_h = offset_bottom_rois[3] * spatial_scale - offset;
T roi_width = roi_end_w - roi_start_w;
T roi_height = roi_end_h - roi_start_h;
if (!half_pixel) {
// Force malformed ROIs to be 1x1
roi_width = std::max(roi_width, (T)1.);
roi_height = std::max(roi_height, (T)1.);
}
T bin_size_h = static_cast<T>(roi_height) / static_cast<T>(pooled_height);
T bin_size_w = static_cast<T>(roi_width) / static_cast<T>(pooled_width);
// We use roi_bin_grid to sample the grid and mimic integral
int64_t roi_bin_grid_h = (sampling_ratio > 0) ? sampling_ratio : static_cast<int64_t>(std::ceil(roi_height / pooled_height)); // e.g., = 2
int64_t roi_bin_grid_w =
(sampling_ratio > 0) ? sampling_ratio : static_cast<int64_t>(std::ceil(roi_width / pooled_width));
// We do average (integral) pooling inside a bin
const int64_t count = std::max(roi_bin_grid_h * roi_bin_grid_w, static_cast<int64_t>(1)); // e.g. = 4
// we want to precalculate indices and weights shared by all channels,
// this is the key point of optimization
std::vector<PreCalc<T>> pre_calc(roi_bin_grid_h * roi_bin_grid_w * pooled_width * SafeInt<size_t>(pooled_height));
PreCalcForBilinearInterpolate(height, width, pooled_height, pooled_width, roi_bin_grid_h, roi_bin_grid_w,
roi_start_h, roi_start_w, bin_size_h, bin_size_w, roi_bin_grid_h,
roi_bin_grid_w, pre_calc);
for (int64_t c = 0; c < channels; c++) {
int64_t index_n_c = index_n + c * pooled_width * pooled_height;
const T* offset_bottom_data =
bottom_data + static_cast<int64_t>((roi_batch_ind * channels + c) * height * width);
int64_t pre_calc_index = 0;
for (int64_t ph = 0; ph < pooled_height; ph++) {
for (int64_t pw = 0; pw < pooled_width; pw++) {
int64_t index = index_n_c + ph * pooled_width + pw;
T output_val = 0.;
if (mode == RoiAlignMode::avg) { // avg pooling
for (int64_t iy = 0; iy < roi_bin_grid_h; iy++) {
for (int64_t ix = 0; ix < roi_bin_grid_w; ix++) {
const auto& pc = pre_calc[onnxruntime::narrow<size_t>(pre_calc_index)];
output_val += pc.w1 * offset_bottom_data[pc.pos1] + pc.w2 * offset_bottom_data[pc.pos2] +
pc.w3 * offset_bottom_data[pc.pos3] + pc.w4 * offset_bottom_data[pc.pos4];
pre_calc_index += 1;
}
}
output_val /= count;
} else { // max pooling
bool max_flag = false;
for (int64_t iy = 0; iy < roi_bin_grid_h; iy++) {
for (int64_t ix = 0; ix < roi_bin_grid_w; ix++) {
const auto& pc = pre_calc[onnxruntime::narrow<size_t>(pre_calc_index)];
T val = std::max(
std::max(std::max(pc.w1 * offset_bottom_data[pc.pos1], pc.w2 * offset_bottom_data[pc.pos2]),
pc.w3 * offset_bottom_data[pc.pos3]),
pc.w4 * offset_bottom_data[pc.pos4]);
if (!max_flag) {
output_val = val;
max_flag = true;
} else {
output_val = std::max(output_val, val);
}
pre_calc_index += 1;
}
}
}
top_data[index] = output_val;
} // for pw
} // for ph
} // for c
} // for n
});
}
} // namespace
Status CheckROIAlignValidInput(const Tensor* X_ptr, const Tensor* rois_ptr, const Tensor* batch_indices_ptr) {
constexpr int64_t EXPECTED_NUM_ROI_DIMS = 2;
constexpr int64_t EXPECTED_SECOND_ROI_DIM = 4;
if (!X_ptr) {
return Status(common::ONNXRUNTIME, common::INVALID_ARGUMENT, "Null input X ptr");
}
if (!rois_ptr) {
return Status(common::ONNXRUNTIME, common::INVALID_ARGUMENT, "Null rois_ptr");
}
if (!batch_indices_ptr) {
return Status(common::ONNXRUNTIME, common::INVALID_ARGUMENT, "Null batch_indices_ptr");
}
const auto& rois_dims = rois_ptr->Shape();
const auto& batch_indices_dims = batch_indices_ptr->Shape();
if (batch_indices_dims.NumDimensions() != 1) {
return Status(common::ONNXRUNTIME, common::INVALID_ARGUMENT,
"Number of dimensions for batch indices should be exactly 1");
}
// validate rois_dims
if (rois_dims.NumDimensions() != EXPECTED_NUM_ROI_DIMS) {
return Status(common::ONNXRUNTIME, common::INVALID_ARGUMENT,
"Number of dimensions for rois should be exactly " + std::to_string(EXPECTED_NUM_ROI_DIMS));
}
if (rois_dims[1] != EXPECTED_SECOND_ROI_DIM) {
return Status(common::ONNXRUNTIME, common::INVALID_ARGUMENT,
"Second dimension for rois should be exactly " + std::to_string(EXPECTED_SECOND_ROI_DIM));
}
// first dimension of batch_indices and rois should match
if (batch_indices_dims[0] != rois_dims[0]) {
return Status(common::ONNXRUNTIME, common::INVALID_ARGUMENT,
"First dimension (num_rois) of batch_indices and rois don't match");
}
return Status::OK();
}
template <typename T>
Status RoiAlign<T>::Compute(OpKernelContext* context) const {
const auto* X_ptr = context->Input<Tensor>(0);
const auto* rois_ptr = context->Input<Tensor>(1);
const auto* batch_indices_ptr = context->Input<Tensor>(2);
const auto& x_dims = X_ptr->Shape();
const auto& rois_dims = rois_ptr->Shape();
const auto& batch_indices_dims = batch_indices_ptr->Shape();
auto num_channels = x_dims[1];
auto num_rois = batch_indices_dims[0];
auto num_roi_cols = rois_dims[1];
auto status = CheckROIAlignValidInput(X_ptr, rois_ptr, batch_indices_ptr);
if (!status.IsOK()) {
return status;
}
auto& Y = *context->Output(0, {num_rois, num_channels, this->output_height_, this->output_width_});
RoiAlignForward<T>(Y.Shape(), X_ptr->Data<T>(), this->spatial_scale_,
x_dims[2], // height
x_dims[3], // width
this->sampling_ratio_, rois_ptr->Data<T>(), num_roi_cols, Y.template MutableData<T>(), this->mode_, this->half_pixel_,
batch_indices_ptr->Data<int64_t>(), context->GetOperatorThreadPool());
return Status::OK();
}
} // namespace onnxruntime