JYLongKong
commented
6 years ago 问了旷视客服，给出了解决方案
`//初始化
long handle1 = NativeFaceppAPI.nativeInit(getApplicationContext(),
ConUtil.getFileContent(getApplicationContext(), R.raw.megviifacepp_0_5_2_model), 0);
//picture是图片的路径，对图片进行转化，这里是转成灰度图
byte[] imageData = FileImageUtils.bitmap2Gray(picture);
int width = FileImageUtils.getSize(picture)[0];
int height = FileImageUtils.getSize(picture)[1];
//这里是设置了默认的参数，有参数变动可以这里修改
float[] configs = NativeFaceppAPI.nativeGetFaceppConfig(handle1);
Assert.assertNotNull(configs);
Facepp.FaceppConfig faceppConfig = new Facepp.FaceppConfig();
faceppConfig.minFaceSize = (int) configs[0];
faceppConfig.rotation = 0;
faceppConfig.interval = (int) configs[2];
faceppConfig.detectionMode = (int) configs[3];
faceppConfig.roi_left = (int) configs[4];
faceppConfig.roi_top = (int) configs[5];
faceppConfig.roi_right = (int) configs[6];
faceppConfig.roi_bottom = (int) configs[7];
faceppConfig.face_confidence_filter=configs[8];
faceppConfig.one_face_tracking = (int) configs[9];
faceppConfig.one_face_tracking = 0;
faceppConfig.detectionMode = Facepp.FaceppConfig.DETECTION_MODE_NORMAL;
NativeFaceppAPI.nativeSetFaceppConfig(handle1, faceppConfig.minFaceSize, faceppConfig.rotation,
faceppConfig.interval, faceppConfig.detectionMode, faceppConfig.roi_left,
faceppConfig.roi_top, faceppConfig.roi_right, faceppConfig.roi_bottom,faceppConfig.face_confidence_filter, faceppConfig.one_face_tracking);
int faceSize = NativeFaceppAPI.nativeDetect(handle1,imageData, width,height,Facepp.IMAGEMODE_GRAY);
if (faceSize <= 0) {
//未被标记出来的人脸
Log.d("facepp","未检出人脸");
}else{
Facepp.Face[] faces = new Facepp.Face[faceSize];
for (int i = 0; i < faceSize; i++) {
Facepp.Face face = new Facepp.Face();
face.points = new PointF[81];
float[] landmark = NativeFaceppAPI.nativeLandMark(handle1, i, Facepp.FPP_GET_LANDMARK81);
float[] points = NativeFaceppAPI.nativeFaceInfo(handle1, i);
for (int j = 0; j < 81; j++) {
PointF point = new PointF();
point.x = (int) landmark[2 * j];
point.y = (int) landmark[2 * j + 1];
face.points[j] = point;
}
faces[i] = face;
face.confidence = points[2];
}
for(int i = 0; i < faces.length; i++){
String detectResult =" face_index: "+i+" confidence=="+faces[i].confidence+" index--"+i;
Log.d("facepp","每个人脸检出结果："+detectResult);
}
//检测后，画关键点，并保存
Bitmap bitmap2 = Utils.drawPoint(faces, picture);
if (bitmap2 != null) {
Utils.saveMyBitmap("sdcard/"+"detect_face","landmark_img.png", bitmap2);
}
}
}`
稍微修改了一下demo，尝试用图片bitmap转YUV格式后作为detect的第一个参数，但是没有发现检测到的人脸