tf.util.encodeString(img64, 'base64').buffer returns empty array

TensorFlow General Discussion
1

Hello,

In my React Native app using tensorflow.js, I retrieve an image (a file) as a base64 type with:

const img64 = await FileSystem.readAsStringAsync(result.uri, {encoding:FileSystem.EncodingType.Base64})

This returns something like this:

iVBORw0KGgoAAAANSUhEUgAAAFoAAABaCAIAAAC3ytZVAAAAAXNSR0IArs4c6QAAAERlWElmTU0AKgAAAAgAAYdpAAQAAAABAAAAGgAAAAAAA6ABAAMAAAABAAEAAKACAAQAAAABAAAAWqADAAQAAAABAAAAWgAAAABJfIu3AAAA......

But then when I call:

const imgBuffer = tf.util.encodeString(img64, 'base64').buffer

it returns an empty array…

3

Hi @aymeric_B ,

I’ve implemented base64 string to buffer conversion in my react-native app using the following code:

const resizedImage = await ImageManipulator.manipulateAsync(
  selectedImage,
  [{ resize: { width: 640, height: 640 } }],
  { format: ImageManipulator.SaveFormat.PNG, base64: true }
);
const resizedImagBase = resizedImage.base64 ??  ''; 
console.log("resizedImage..", resizedImagBase)
const imgBuffer = tf.util.encodeString( resizedImagBase
  , 'base64').buffer;
console.log("imgBuffer",imgBuffer)

Output:

image
image1890×526 64.3 KB

Let me know if helps. Thank You!!