screen: update tresholds on text

app.go

@@ -0,0 +1,61 @@
+package goadb
+
+import (
+	"fmt"
+	"regexp"
+	"strconv"
+	"strings"
+)
+
+// Get the current focussed app package name
+func (d *Device) GetCurrentAppPackage() (string, error) {
+	data, err := d.RunCommand("shell dumpsys activity activities | grep ResumedActivity")
+	re := regexp.MustCompile(`\b([a-zA-Z0-9._]+)/(?:[a-zA-Z0-9._]+)\b`)
+	matches := re.FindStringSubmatch(string(data))
+
+	if len(matches) > 1 {
+		return matches[1], nil
+	} else {
+		return "", nil
+	}
+	return "", err
+}
+
+// Get the pid for a app
+func (d *Device) GetAppPid(appackage string) (int, error) {
+	data, err := d.RunCommand(fmt.Sprintf("shell pidof %s"))
+	if err != nil {
+		return -1, err
+	}
+
+	pid, err := strconv.Atoi(strings.TrimSpace(string(data)))
+	if err != nil {
+		return -1, err
+	}
+
+	return pid, nil
+}
+
+// Check if a app is running
+func (d *Device) CheckAppRunning(appackage string) (bool, error) {
+	pid, err := d.GetAppPid(appackage)
+	if err != nil {
+		return false, err
+	}
+	if pid != -1 {
+		return true, nil
+	}
+	return false, nil
+}
+
+// Start a app using the package name
+func (d *Device) StartApp(appackage string) error {
+	_, err := d.RunCommand(fmt.Sprintf("shell monkey -p %s -c android.intent.category.LAUNCHER 1", appackage))
+	return err
+}
+
+// Force kill a running app
+func (d *Device) KillApp(appackage string) error {
+	_, err := d.RunCommand(fmt.Sprintf("shell am force-stop %s", appackage))
+	return err
+}

screen.go

@@ -28,6 +28,12 @@ func (d *Device) ToggleScreen() error {
 	return err
 }
 
+// Tap on a location on the screen
+func (d *Device) TapScreen(x int, y int) error {
+	_, err := d.RunCommand(fmt.Sprintf("shell input tap %d %d", x, y))
+	return err
+}
+
 // Get the resolution of the screen
 func (d *Device) GetScreenResolution() (int, int, error) {
 
@@ -61,21 +67,17 @@ func (d *Device) GetScreenResolution() (int, int, error) {
 	return d.Screenx, d.Screeny, nil
 }
 
-// Get the text on screen in a certain area using gotesseract
-func (d *Device) GetScreenText(xStart int, xEnd int, yStart int, yEnd int, whitelist string, multiline bool) (string, error) {
-	// Take screenshot in memory and get bytes
+func (d *Device) GetScreenText(xStart int, xEnd int, yStart int, yEnd int, whitelist string, multiline bool, textPolarity string) (string, error) {
 	data, err := d.RunCommand("exec-out screencap -p")
 	if err != nil {
 		return "", fmt.Errorf("failed to take screenshot: %w", err)
 	}
 
-	// Decode image from bytes
 	img, err := png.Decode(bytes.NewReader(data))
 	if err != nil {
 		return "", fmt.Errorf("failed to decode PNG: %w", err)
 	}
 
-	// Crop region
 	rect := image.Rect(0, 0, xEnd-xStart, yEnd-yStart)
 	cropped := image.NewRGBA(rect)
 	for y := yStart; y < yEnd; y++ {
@@ -84,27 +86,57 @@ func (d *Device) GetScreenText(xStart int, xEnd int, yStart int, yEnd int, white
 		}
 	}
 
-	// Convert to grayscale and apply threshold
-	threshold := uint8(160)
+	// Convert to grayscale and compute average brightness
+	gray := image.NewGray(rect)
+	var totalBrightness int
+	for y := 0; y < rect.Dy(); y++ {
+		for x := 0; x < rect.Dx(); x++ {
+			grayVal := color.GrayModel.Convert(cropped.At(x, y)).(color.Gray)
+			gray.Set(x, y, grayVal)
+			totalBrightness += int(grayVal.Y)
+		}
+	}
+	avgBrightness := totalBrightness / (rect.Dx() * rect.Dy())
+
+	// Decide polarity
+	useLightText := false
+	switch textPolarity {
+	case "light":
+		useLightText = true
+	case "dark":
+		useLightText = false
+	case "auto":
+		useLightText = avgBrightness < 128
+	default:
+		return "", fmt.Errorf("invalid textPolarity: must be 'dark', 'light', or 'auto'")
+	}
+
+	// Binarize based on polarity
 	binarized := image.NewGray(rect)
 	for y := 0; y < rect.Dy(); y++ {
 		for x := 0; x < rect.Dx(); x++ {
-			gray := color.GrayModel.Convert(cropped.At(x, y)).(color.Gray)
-			if gray.Y > threshold {
-				binarized.SetGray(x, y, color.Gray{Y: 255})
+			grayVal := gray.GrayAt(x, y).Y
+			if useLightText {
+				if int(grayVal) > avgBrightness+10 {
+					binarized.SetGray(x, y, color.Gray{Y: 0})
+				} else {
+					binarized.SetGray(x, y, color.Gray{Y: 255})
+				}
 			} else {
-				binarized.SetGray(x, y, color.Gray{Y: 0})
+				if int(grayVal) < avgBrightness-10 {
+					binarized.SetGray(x, y, color.Gray{Y: 0})
+				} else {
+					binarized.SetGray(x, y, color.Gray{Y: 255})
+				}
 			}
 		}
 	}
 
-	// Encode binarized image to PNG
 	var buf bytes.Buffer
 	if err := png.Encode(&buf, binarized); err != nil {
 		return "", fmt.Errorf("failed to encode image: %w", err)
 	}
 
-	// Call tesseract
 	client := gosseract.NewClient()
 	defer client.Close()
 	client.SetImageFromBytes(buf.Bytes())