haarobjectdetect_scaled2.cl 13.6 KB
Newer Older
wester committed
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
/*M///////////////////////////////////////////////////////////////////////////////////////
//
//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
//
//  By downloading, copying, installing or using the software you agree to this license.
//  If you do not agree to this license, do not download, install,
//  copy or use the software.
//
//
//                           License Agreement
//                For Open Source Computer Vision Library
//
// Copyright (C) 2010-2012, Institute Of Software Chinese Academy Of Science, all rights reserved.
// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
// Third party copyrights are property of their respective owners.
//
// @Authors
//    Wu Xinglong, wxl370@126.com
//    Sen Liu, swjtuls1987@126.com
//    Peng Xiao, pengxiao@outlook.com
//    Erping Pang, erping@multicorewareinc.com
// Redistribution and use in source and binary forms, with or without modification,
// are permitted provided that the following conditions are met:
//
//   * Redistribution's of source code must retain the above copyright notice,
//     this list of conditions and the following disclaimer.
//
//   * Redistribution's in binary form must reproduce the above copyright notice,
//     this list of conditions and the following disclaimer in the documentation
//     and/or other materials provided with the distribution.
//
//   * The name of the copyright holders may not be used to endorse or promote products
//     derived from this software without specific prior written permission.
//
// This software is provided by the copyright holders and contributors as is and
// any express or implied warranties, including, but not limited to, the implied
// warranties of merchantability and fitness for a particular purpose are disclaimed.
// In no event shall the Intel Corporation or contributors be liable for any direct,
// indirect, incidental, special, exemplary, or consequential damages
// (including, but not limited to, procurement of substitute goods or services;
// loss of use, data, or profits; or business interruption) however caused
// and on any theory of liability, whether in contract, strict liability,
// or tort (including negligence or otherwise) arising in any way out of
// the use of this software, even if advised of the possibility of such damage.
//
//M*/

#define CV_HAAR_FEATURE_MAX           3
typedef int   sumtype;
typedef float sqsumtype;

typedef struct __attribute__((aligned(128))) GpuHidHaarTreeNode
{
    int p[CV_HAAR_FEATURE_MAX][4] __attribute__((aligned(64)));
    float weight[CV_HAAR_FEATURE_MAX] /*__attribute__((aligned (16)))*/;
    float threshold /*__attribute__((aligned (4)))*/;
    float alpha[3] __attribute__((aligned(16)));
    int left __attribute__((aligned(4)));
    int right __attribute__((aligned(4)));
}
GpuHidHaarTreeNode;
typedef struct __attribute__((aligned(32))) GpuHidHaarClassifier
{
    int count __attribute__((aligned(4)));
    GpuHidHaarTreeNode *node __attribute__((aligned(8)));
    float *alpha __attribute__((aligned(8)));
}
GpuHidHaarClassifier;
typedef struct __attribute__((aligned(64))) GpuHidHaarStageClassifier
{
    int  count __attribute__((aligned(4)));
    float threshold __attribute__((aligned(4)));
    int two_rects __attribute__((aligned(4)));
    int reserved0 __attribute__((aligned(8)));
    int reserved1 __attribute__((aligned(8)));
    int reserved2 __attribute__((aligned(8)));
    int reserved3 __attribute__((aligned(8)));
}
GpuHidHaarStageClassifier;
typedef struct __attribute__((aligned(64))) GpuHidHaarClassifierCascade
{
    int  count __attribute__((aligned(4)));
    int  is_stump_based __attribute__((aligned(4)));
    int  has_tilted_features __attribute__((aligned(4)));
    int  is_tree __attribute__((aligned(4)));
    int pq0 __attribute__((aligned(4)));
    int pq1 __attribute__((aligned(4)));
    int pq2 __attribute__((aligned(4)));
    int pq3 __attribute__((aligned(4)));
    int p0 __attribute__((aligned(4)));
    int p1 __attribute__((aligned(4)));
    int p2 __attribute__((aligned(4)));
    int p3 __attribute__((aligned(4)));
    float inv_window_area __attribute__((aligned(4)));
} GpuHidHaarClassifierCascade;

__kernel void gpuRunHaarClassifierCascade_scaled2(
    global GpuHidHaarStageClassifier *stagecascadeptr,
    global int4 *info,
    global GpuHidHaarTreeNode *nodeptr,
    global const int *restrict sum,
    global const float   *restrict sqsum,
    global int4 *candidate,
    const int rows,
    const int cols,
    const int step,
    const int loopcount,
    const int start_stage,
    const int split_stage,
    const int end_stage,
    const int startnode,
    global int4 *p,
    global float *correction,
    const int nodecount)
{
    int grpszx = get_local_size(0);
    int grpszy = get_local_size(1);
    int grpnumx = get_num_groups(0);
    int grpidx = get_group_id(0);
    int lclidx = get_local_id(0);
    int lclidy = get_local_id(1);
    int lcl_id = mad24(lclidy, grpszx, lclidx);
    __local int glboutindex[1];
    __local int lclcount[1];
    __local int lcloutindex[64];
    glboutindex[0] = 0;
    int outputoff = mul24(grpidx, 256);
    candidate[outputoff + (lcl_id << 2)] = (int4)0;
    candidate[outputoff + (lcl_id << 2) + 1] = (int4)0;
    candidate[outputoff + (lcl_id << 2) + 2] = (int4)0;
    candidate[outputoff + (lcl_id << 2) + 3] = (int4)0;
    int max_idx = rows * cols - 1;
    for (int scalei = 0; scalei < loopcount; scalei++)
    {
        int4 scaleinfo1;
        scaleinfo1 = info[scalei];
        int grpnumperline = (scaleinfo1.y & 0xffff0000) >> 16;
        int totalgrp = scaleinfo1.y & 0xffff;
        float factor = as_float(scaleinfo1.w);
        float correction_t = correction[scalei];
        float ystep = max(2.0f, factor);

        for (int grploop = get_group_id(0); grploop < totalgrp; grploop += grpnumx)
        {
            int4 cascadeinfo = p[scalei];
            int grpidy = grploop / grpnumperline;
            int grpidx = grploop - mul24(grpidy, grpnumperline);
            int ix = mad24(grpidx, grpszx, lclidx);
            int iy = mad24(grpidy, grpszy, lclidy);
            int x = round(ix * ystep);
            int y = round(iy * ystep);
            lcloutindex[lcl_id] = 0;
            lclcount[0] = 0;
            int nodecounter;
            float mean, variance_norm_factor;
            //if((ix < width) && (iy < height))
            {
                const int p_offset = mad24(y, step, x);
                cascadeinfo.x += p_offset;
                cascadeinfo.z += p_offset;
                mean = (sum[clamp(mad24(cascadeinfo.y, step, cascadeinfo.x), 0, max_idx)]
                - sum[clamp(mad24(cascadeinfo.y, step, cascadeinfo.z), 0, max_idx)] -
                        sum[clamp(mad24(cascadeinfo.w, step, cascadeinfo.x), 0, max_idx)]
                + sum[clamp(mad24(cascadeinfo.w, step, cascadeinfo.z), 0, max_idx)])
                       * correction_t;
                variance_norm_factor = sqsum[clamp(mad24(cascadeinfo.y, step, cascadeinfo.x), 0, max_idx)]
                - sqsum[clamp(mad24(cascadeinfo.y, step, cascadeinfo.z), 0, max_idx)] -
                                       sqsum[clamp(mad24(cascadeinfo.w, step, cascadeinfo.x), 0, max_idx)]
                + sqsum[clamp(mad24(cascadeinfo.w, step, cascadeinfo.z), 0, max_idx)];
                variance_norm_factor = variance_norm_factor * correction_t - mean * mean;
                variance_norm_factor = variance_norm_factor >= 0.f ? sqrt(variance_norm_factor) : 1.f;
                bool result = true;
                nodecounter = startnode + nodecount * scalei;
                for (int stageloop = start_stage; (stageloop < end_stage) && result; stageloop++)
                {
                    float stage_sum = 0.f;
                    int   stagecount = stagecascadeptr[stageloop].count;
                    for (int nodeloop = 0; nodeloop < stagecount;)
                    {
                        __global GpuHidHaarTreeNode *currentnodeptr = (nodeptr + nodecounter);
                        int4 info1 = *(__global int4 *)(&(currentnodeptr->p[0][0]));
                        int4 info2 = *(__global int4 *)(&(currentnodeptr->p[1][0]));
                        int4 info3 = *(__global int4 *)(&(currentnodeptr->p[2][0]));
                        float4 w = *(__global float4 *)(&(currentnodeptr->weight[0]));
                        float3 alpha3 = *(__global float3 *)(&(currentnodeptr->alpha[0]));
                        float nodethreshold  = w.w * variance_norm_factor;

                        info1.x += p_offset;
                        info1.z += p_offset;
                        info2.x += p_offset;
                        info2.z += p_offset;
                        info3.x += p_offset;
                        info3.z += p_offset;
                        float classsum = (sum[clamp(mad24(info1.y, step, info1.x), 0, max_idx)]
                        - sum[clamp(mad24(info1.y, step, info1.z), 0, max_idx)] -
                                          sum[clamp(mad24(info1.w, step, info1.x), 0, max_idx)]
                        + sum[clamp(mad24(info1.w, step, info1.z), 0, max_idx)]) * w.x;
                        classsum += (sum[clamp(mad24(info2.y, step, info2.x), 0, max_idx)]
                        - sum[clamp(mad24(info2.y, step, info2.z), 0, max_idx)] -
                                     sum[clamp(mad24(info2.w, step, info2.x), 0, max_idx)]
                        + sum[clamp(mad24(info2.w, step, info2.z), 0, max_idx)]) * w.y;
                        classsum += (sum[clamp(mad24(info3.y, step, info3.x), 0, max_idx)]
                        - sum[clamp(mad24(info3.y, step, info3.z), 0, max_idx)] -
                                     sum[clamp(mad24(info3.w, step, info3.x), 0, max_idx)]
                        + sum[clamp(mad24(info3.w, step, info3.z), 0, max_idx)]) * w.z;

                        bool passThres = classsum >= nodethreshold;

#if STUMP_BASED
                        stage_sum += passThres ? alpha3.y : alpha3.x;
                        nodecounter++;
                        nodeloop++;
#else
                        bool isRootNode = (nodecounter & 1) == 0;
                        if(isRootNode)
                        {
                            if( (passThres && currentnodeptr->right) ||
                                (!passThres && currentnodeptr->left))
                            {
                                nodecounter ++;
                            }
                            else
                            {
                                stage_sum += alpha3.x;
                                nodecounter += 2;
                                nodeloop ++;
                            }
                        }
                        else
                        {
                            stage_sum += (passThres ? alpha3.z : alpha3.y);
                            nodecounter ++;
                            nodeloop ++;
                        }
#endif
                    }
                    result = (int)(stage_sum >= stagecascadeptr[stageloop].threshold);
                }

                barrier(CLK_LOCAL_MEM_FENCE);

                if (result)
                {
                    int queueindex = atomic_inc(lclcount);
                    lcloutindex[queueindex] = (y << 16) | x;
                }
                barrier(CLK_LOCAL_MEM_FENCE);
                int queuecount = lclcount[0];

                if (lcl_id < queuecount)
                {
                    int temp = lcloutindex[lcl_id];
                    int x = temp & 0xffff;
                    int y = (temp & (int)0xffff0000) >> 16;
                    temp = atomic_inc(glboutindex);
                    int4 candidate_result;
                    candidate_result.zw = (int2)convert_int_rte(factor * 20.f);
                    candidate_result.x = x;
                    candidate_result.y = y;

                    int i = outputoff+temp+lcl_id;
                    if(candidate[i].z == 0)
                    {
                        candidate[i] = candidate_result;
                    }
                    else
                    {
                        for(i=i+1;;i++)
                        {
                            if(candidate[i].z == 0)
                            {
                                candidate[i] = candidate_result;
                                break;
                            }
                        }
                    }
                }

                barrier(CLK_LOCAL_MEM_FENCE);
            }
        }
    }
}
__kernel void gpuscaleclassifier(global GpuHidHaarTreeNode *orinode, global GpuHidHaarTreeNode *newnode, float scale, float weight_scale, int nodenum)
{
    int counter = get_global_id(0);
    int tr_x[3], tr_y[3], tr_h[3], tr_w[3], i = 0;
    GpuHidHaarTreeNode t1 = *(orinode + counter);

    #pragma unroll
    for (i = 0; i < 3; i++)
    {
        tr_x[i] = (int)(t1.p[i][0] * scale + 0.5f);
        tr_y[i] = (int)(t1.p[i][1] * scale + 0.5f);
        tr_w[i] = (int)(t1.p[i][2] * scale + 0.5f);
        tr_h[i] = (int)(t1.p[i][3] * scale + 0.5f);
    }

    t1.weight[0] = -(t1.weight[1] * tr_h[1] * tr_w[1] + t1.weight[2] * tr_h[2] * tr_w[2]) / (tr_h[0] * tr_w[0]);
    counter += nodenum;

    #pragma unroll
    for (i = 0; i < 3; i++)
    {
        newnode[counter].p[i][0] = tr_x[i];
        newnode[counter].p[i][1] = tr_y[i];
        newnode[counter].p[i][2] = tr_x[i] + tr_w[i];
        newnode[counter].p[i][3] = tr_y[i] + tr_h[i];
        newnode[counter].weight[i] = t1.weight[i] * weight_scale;
    }

    newnode[counter].left = t1.left;
    newnode[counter].right = t1.right;
    newnode[counter].threshold = t1.threshold;
    newnode[counter].alpha[0] = t1.alpha[0];
    newnode[counter].alpha[1] = t1.alpha[1];
    newnode[counter].alpha[2] = t1.alpha[2];
}