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
|
/*
* Copyright (c) 2019 Xuewei Meng
*
* This file is part of FFmpeg.
*
* FFmpeg is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
* FFmpeg is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with FFmpeg; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
/**
* @file
* Filter implementing image derain filter using deep convolutional networks.
* http://openaccess.thecvf.com/content_ECCV_2018/html/Xia_Li_Recurrent_Squeeze-and-Excitation_Context_ECCV_2018_paper.html
*/
#include "libavformat/avio.h"
#include "libavutil/opt.h"
#include "avfilter.h"
#include "dnn_interface.h"
#include "formats.h"
#include "internal.h"
typedef struct DRContext {
const AVClass *class;
int filter_type;
char *model_filename;
DNNBackendType backend_type;
DNNModule *dnn_module;
DNNModel *model;
DNNData input;
DNNData output;
} DRContext;
#define CLIP(x, min, max) (x < min ? min : (x > max ? max : x))
#define OFFSET(x) offsetof(DRContext, x)
#define FLAGS AV_OPT_FLAG_FILTERING_PARAM | AV_OPT_FLAG_VIDEO_PARAM
static const AVOption derain_options[] = {
{ "filter_type", "filter type(derain/dehaze)", OFFSET(filter_type), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 1, FLAGS, "type" },
{ "derain", "derain filter flag", 0, AV_OPT_TYPE_CONST, { .i64 = 0 }, 0, 0, FLAGS, "type" },
{ "dehaze", "dehaze filter flag", 0, AV_OPT_TYPE_CONST, { .i64 = 1 }, 0, 0, FLAGS, "type" },
{ "dnn_backend", "DNN backend", OFFSET(backend_type), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 1, FLAGS, "backend" },
{ "native", "native backend flag", 0, AV_OPT_TYPE_CONST, { .i64 = 0 }, 0, 0, FLAGS, "backend" },
#if (CONFIG_LIBTENSORFLOW == 1)
{ "tensorflow", "tensorflow backend flag", 0, AV_OPT_TYPE_CONST, { .i64 = 1 }, 0, 0, FLAGS, "backend" },
#endif
{ "model", "path to model file", OFFSET(model_filename), AV_OPT_TYPE_STRING, { .str = NULL }, 0, 0, FLAGS },
{ NULL }
};
AVFILTER_DEFINE_CLASS(derain);
static int query_formats(AVFilterContext *ctx)
{
AVFilterFormats *formats;
const enum AVPixelFormat pixel_fmts[] = {
AV_PIX_FMT_RGB24,
AV_PIX_FMT_NONE
};
formats = ff_make_format_list(pixel_fmts);
return ff_set_common_formats(ctx, formats);
}
static int config_inputs(AVFilterLink *inlink)
{
AVFilterContext *ctx = inlink->dst;
DRContext *dr_context = ctx->priv;
const char *model_output_name = "y";
DNNReturnType result;
dr_context->input.width = inlink->w;
dr_context->input.height = inlink->h;
dr_context->input.channels = 3;
result = (dr_context->model->set_input_output)(dr_context->model->model, &dr_context->input, "x", &model_output_name, 1);
if (result != DNN_SUCCESS) {
av_log(ctx, AV_LOG_ERROR, "could not set input and output for the model\n");
return AVERROR(EIO);
}
return 0;
}
static int filter_frame(AVFilterLink *inlink, AVFrame *in)
{
AVFilterContext *ctx = inlink->dst;
AVFilterLink *outlink = ctx->outputs[0];
DRContext *dr_context = ctx->priv;
DNNReturnType dnn_result;
int pad_size;
AVFrame *out = ff_get_video_buffer(outlink, outlink->w, outlink->h);
if (!out) {
av_log(ctx, AV_LOG_ERROR, "could not allocate memory for output frame\n");
av_frame_free(&in);
return AVERROR(ENOMEM);
}
av_frame_copy_props(out, in);
for (int i = 0; i < in->height; i++){
for(int j = 0; j < in->width * 3; j++){
int k = i * in->linesize[0] + j;
int t = i * in->width * 3 + j;
((float *)dr_context->input.data)[t] = in->data[0][k] / 255.0;
}
}
dnn_result = (dr_context->dnn_module->execute_model)(dr_context->model, &dr_context->output, 1);
if (dnn_result != DNN_SUCCESS){
av_log(ctx, AV_LOG_ERROR, "failed to execute model\n");
return AVERROR(EIO);
}
out->height = dr_context->output.height;
out->width = dr_context->output.width;
outlink->h = dr_context->output.height;
outlink->w = dr_context->output.width;
pad_size = (in->height - out->height) >> 1;
for (int i = 0; i < out->height; i++){
for(int j = 0; j < out->width * 3; j++){
int k = i * out->linesize[0] + j;
int t = i * out->width * 3 + j;
int t_in = (i + pad_size) * in->width * 3 + j + pad_size * 3;
out->data[0][k] = CLIP((int)((((float *)dr_context->input.data)[t_in] - ((float *)dr_context->output.data)[t]) * 255), 0, 255);
}
}
av_frame_free(&in);
return ff_filter_frame(outlink, out);
}
static av_cold int init(AVFilterContext *ctx)
{
DRContext *dr_context = ctx->priv;
dr_context->input.dt = DNN_FLOAT;
dr_context->dnn_module = ff_get_dnn_module(dr_context->backend_type);
if (!dr_context->dnn_module) {
av_log(ctx, AV_LOG_ERROR, "could not create DNN module for requested backend\n");
return AVERROR(ENOMEM);
}
if (!dr_context->model_filename) {
av_log(ctx, AV_LOG_ERROR, "model file for network is not specified\n");
return AVERROR(EINVAL);
}
if (!dr_context->dnn_module->load_model) {
av_log(ctx, AV_LOG_ERROR, "load_model for network is not specified\n");
return AVERROR(EINVAL);
}
dr_context->model = (dr_context->dnn_module->load_model)(dr_context->model_filename);
if (!dr_context->model) {
av_log(ctx, AV_LOG_ERROR, "could not load DNN model\n");
return AVERROR(EINVAL);
}
return 0;
}
static av_cold void uninit(AVFilterContext *ctx)
{
DRContext *dr_context = ctx->priv;
if (dr_context->dnn_module) {
(dr_context->dnn_module->free_model)(&dr_context->model);
av_freep(&dr_context->dnn_module);
}
}
static const AVFilterPad derain_inputs[] = {
{
.name = "default",
.type = AVMEDIA_TYPE_VIDEO,
.config_props = config_inputs,
.filter_frame = filter_frame,
},
{ NULL }
};
static const AVFilterPad derain_outputs[] = {
{
.name = "default",
.type = AVMEDIA_TYPE_VIDEO,
},
{ NULL }
};
AVFilter ff_vf_derain = {
.name = "derain",
.description = NULL_IF_CONFIG_SMALL("Apply derain filter to the input."),
.priv_size = sizeof(DRContext),
.init = init,
.uninit = uninit,
.query_formats = query_formats,
.inputs = derain_inputs,
.outputs = derain_outputs,
.priv_class = &derain_class,
.flags = AVFILTER_FLAG_SUPPORT_TIMELINE_GENERIC,
};
|
