ncnn

1
// Tencent is pleased to support the open source community by making ncnn available.
2
//
3
// Copyright (C) 2020 THL A29 Limited, a Tencent company. All rights reserved.
4
//
5
// Licensed under the BSD 3-Clause License (the "License"); you may not use this file except
6
// in compliance with the License. You may obtain a copy of the License at
7
//
8
// https://opensource.org/licenses/BSD-3-Clause
9
//
10
// Unless required by applicable law or agreed to in writing, software distributed
11
// under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR
12
// CONDITIONS OF ANY KIND, either express or implied. See the License for the
13
// specific language governing permissions and limitations under the License.
14

15
#include "pixelshuffle.h"
16

17
namespace ncnn {
18

19
PixelShuffle::PixelShuffle()
20
{
21
    one_blob_only = true;
22
    support_inplace = false;
23
}
24

25
int PixelShuffle::load_param(const ParamDict& pd)
26
{
27
    upscale_factor = pd.get(0, 1);
28
    mode = pd.get(1, 0);
29

30
    return 0;
31
}
32

33
int PixelShuffle::forward(const Mat& bottom_blob, Mat& top_blob, const Option& opt) const
34
{
35
    int w = bottom_blob.w;
36
    int h = bottom_blob.h;
37
    int channels = bottom_blob.c;
38
    size_t elemsize = bottom_blob.elemsize;
39

40
    int outw = w * upscale_factor;
41
    int outh = h * upscale_factor;
42
    int outc = channels / (upscale_factor * upscale_factor);
43

44
    top_blob.create(outw, outh, outc, elemsize, opt.blob_allocator);
45
    if (top_blob.empty())
46
        return -100;
47

48
    #pragma omp parallel for num_threads(opt.num_threads)
49
    for (int p = 0; p < outc; p++)
50
    {
51
        Mat m = top_blob.channel(p);
52

53
        for (int sh = 0; sh < upscale_factor; sh++)
54
        {
55
            for (int sw = 0; sw < upscale_factor; sw++)
56
            {
57
                int q;
58
                if (mode == 0)
59
                    q = p * upscale_factor * upscale_factor + sh * upscale_factor + sw;
60
                else // if (mode == 1)
61
                    q = (sh * upscale_factor + sw) * outc + p;
62

63
                const float* sptr = bottom_blob.channel(q);
64

65
                for (int i = 0; i < h; i++)
66
                {
67
                    float* outptr = m.row(i * upscale_factor + sh) + sw;
68
                    for (int j = 0; j < w; j++)
69
                    {
70
                        outptr[0] = sptr[0];
71

72
                        sptr++;
73
                        outptr += upscale_factor;
74
                    }
75
                }
76
            }
77
        }
78
    }
79

80
    return 0;
81
}
82

83
} // namespace ncnn
84