feat: optimize yoloface preprocess

model/util.go

@@ -1,37 +1,23 @@
 package model
 
 import (
-	"bytes"
-	"encoding/binary"
+	"errors"
 	"image"
 	"image/color"
-	"io"
 	"math"
+	"unsafe"
 
 	"gocv.io/x/gocv"
 )
 
 func BytesToFloat32Slice(data []byte) ([]float32, error) {
-	t := []float32{}
-
-	// Create a buffer from the input data
-	buffer := bytes.NewReader(data)
-	for {
-		// Read the binary data from the buffer
-		var binaryValue uint32
-		err := binary.Read(buffer, binary.LittleEndian, &binaryValue)
-		if err != nil {
-			if err == io.EOF {
-				break
-			}
-			return nil, err
-		}
-
-		t = append(t, math.Float32frombits(binaryValue))
-
+	if len(data)%4 != 0 {
+		return nil, errors.New("byte slice length must be divisible by 4")
 	}
 
-	return t, nil
+	floatCount := len(data) / 4
+	floatSlice := unsafe.Slice((*float32)(unsafe.Pointer(&data[0])), floatCount)
+	return floatSlice, nil
 }
 
 // Argmax return the index of the maximum value in a slice
@@ -151,17 +137,20 @@ func MatSubtract(mat gocv.Mat, v float64) {
 // ClipMat clips the values of a gocv.Mat within a specified range.
 // mat need to be float32 type
 func ClipMat(mat gocv.Mat, minVal, maxVal float32) {
-	for row := 0; row < mat.Rows(); row++ {
-		for col := 0; col < mat.Cols(); col++ {
-			value := mat.GetFloatAt(row, col)
-			if value < minVal {
-				value = minVal
-			} else if value > maxVal {
-				value = maxVal
-			}
-			mat.SetFloatAt(row, col, value)
-		}
-	}
+	scalarMin := gocv.NewMatWithSizeFromScalar(
+		gocv.NewScalar(float64(minVal), 0, 0, 0),
+		mat.Rows(), mat.Cols(), mat.Type(),
+	)
+	defer scalarMin.Close()
+
+	scalarMax := gocv.NewMatWithSizeFromScalar(
+		gocv.NewScalar(float64(maxVal), 0, 0, 0),
+		mat.Rows(), mat.Cols(), mat.Type(),
+	)
+	defer scalarMax.Close()
+
+	gocv.Max(mat, scalarMin, &mat)
+	gocv.Min(mat, scalarMax, &mat)
 }
 
 // Padding represents the padding values for the mask.

model/yoloface/pre.go

@@ -64,17 +64,22 @@ func prepareDetectFrame(visionFrame gocv.Mat, faceDetectorSize Resolution) []flo
 	defer roi.Close()
 	visionFrame.CopyTo(&roi)
 
-	output := make([]float32, 3*faceDetectorHeight*faceDetectorWidth)
-	idx := 0
-	for y := 0; y < faceDetectorHeight; y++ {
-		for x := 0; x < faceDetectorWidth; x++ {
-			pixel := detectVisionFrame.GetVecbAt(y, x)
+	data, err := detectVisionFrame.DataPtrUint8()
+	if err != nil {
+		panic(err)
+	}
 
-			output[idx] = (float32(pixel[0]) - 127.5) / 128.0
-			output[faceDetectorHeight*faceDetectorWidth+idx] = (float32(pixel[1]) - 127.5) / 128.0
-			output[2*faceDetectorHeight*faceDetectorWidth+idx] = (float32(pixel[2]) - 127.5) / 128.0
-			idx++
-		}
+	output := make([]float32, 3*faceDetectorHeight*faceDetectorWidth)
+	hw := faceDetectorHeight * faceDetectorWidth
+
+	for i := 0; i < hw; i++ {
+		b := data[i*3]
+		g := data[i*3+1]
+		r := data[i*3+2]
+
+		output[i] = (float32(b) - 127.5) / 128.0
+		output[hw+i] = (float32(g) - 127.5) / 128.0
+		output[2*hw+i] = (float32(r) - 127.5) / 128.0
 	}
 
 	return output