Face recognition model building process

swzswz

Face recognition model building process [Copy link]

 

Face recognition model

Face recognition model steps

Face recognition is mainly divided into two models:

• Detection network model

• Classification Network Model

Model building process

[attach ]650136[/attach]

Video Recording Conversion

Modify the corresponding parameters according to the store_sample routine in the HiSilicon development document . Record the video or use a mobile phone to record the video.

After compiling, use NFS to copy the video on the board to the local computer.

Build the FFmpeg environment and crop the video into pictures.

Dataset creation and annotation

Use labelme tool to label the dataset.

Model Training

We use HiSilicon servers to perform cloud model training using CUDA.

Model Conversion

Generally, pytorch-caffe is used for conversion.

Model Quantization

Convert the caffe model into a wk model for board-side reasoning.

Model deployment and debugging

Video Demonstration

1

Code and other files

static HI_S32 CnnTrashClassifyFlag(const RecogNumInfo items[], HI_S32 itemNum, HI_CHAR* buf, HI_S32 size)
{
   HI_S32 offset = 0;
   HI_CHAR *trashName = NULL;

   offset += snprintf_s(buf + offset, size - offset, size - offset - 1, "face_classify: {");
   for (HI_U32 i = 0; i < itemNum; i++) {
     const RecogNumInfo *item = &items ;
     uint32_t score = item->score * HI_PER_BASE / SCORE_MAX;
     if (score < THRESH_MIN) {
       break;
     }
     SAMPLE_PRT("----trash item flag----num:%d, score:%d\n", item->num, score);
     switch (item->num) {
       case 0u:
         trashName = "face";
         UartSendRead(uartFd1,face);
         SAMPLE_PRT("----trash name----:%s\n", trashName);
         break;
       case 1u:
         trashName = "background";
         UartSendRead(uartFd1,background);
         SAMPLE_PRT("----trash name----:%s\n", trashName);
         break;
       default:
         trashName = "unknown";
         UartSendRead(uartFd1,unknown);
         SAMPLE_PRT("----trash name----:%s\n", trashName);
         break;
     }
     offset += snprintf_s(buf + offset, size - offset, size - offset - 1,
       "%s%s %u:%u%%", (i == 0 ? " " : ", "), trashName, (int)item->num, (int)score);
     HI_ASSERT(offset < size);
   }
   offset += snprintf_s(buf + offset, size - offset, size - offset - 1, " }");
   HI_ASSERT(offset < size);
   return HI_SUCCESS;
}

lugl4313820

Looks very classy, thanks for sharing!

swzswz

lugl4313820 posted on 2022-10-21 22:26 It looks very high-end, thanks for sharing!

I used the HiSilicon development kit and it worked pretty well!