Home/ Questions/Q 2821
Next
Answered
Hiba Zyati
  • 1
Ye Yoooo!
Asked: 2024-11-21T15:07:33+00:00 In:

Using Google's generative AI for images in Kotlin

  • 1

I’m trying to use Google’s generative AI in Kotlin in Android Studio, to write an app that gives me information about a photo. So I would like to pass an image and a text to the gemini-1.5-flash API, but I am not succeeding. This is my code:

            lifecycleScope.launch {
                val prompt = arrayOf(
                    Content.Image(selectedImageBitmap),
                    "What is in this photo?")
                val response = withContext(Dispatchers.IO) {
                    generativeModel.generateContent(prompt)
                }
                withContext(Dispatchers.Main) {
                    OutputValue = response.text
                    myTextOutput.text = OutputValue
                }
            }

selectedImageBitmap is a bitmap object.

In build.grandle.kts file, I’ve added the dependencies:

implementation("com.google.ai.client.generativeai:generativeai:0.9.0")

And in my MainActivity, I’ve added the import:

import com.google.ai.client.generativeai.GenerativeModel

I have the following error: Unresolved reference: Content.

That means, the compiler cannot find the definition of ‘Content.

What’s wrong with my code? What do I need to do to get it to work?

I expected generativeModel.generateContent() to easily accept text, images, PDFs, and other input types.

coding
Share
Leave an answer

Leave an answer

Browse

1 Answer

  1. Best Answer
    God Level!

    The error you’re encountering is because the Content class you’re using isn’t part of the official Google AI Client SDK for Kotlin. Here’s how to fix it and use Google’s generative AI for image analysis in your Kotlin code:

    The Problem:

    The Content class seems to be a custom class or one from a different library. The official Google AI Client SDK doesn’t have a built-in Content class for representing different input types like images.

    The Solution:

    There are two ways to address this:

    1. Using InputImage:

    The Google AI Client SDK offers an InputImage class specifically designed for passing images to the generateContent function. Here’s the corrected code:

    Kotlin
    lifecycleScope.launch {
    val prompt = arrayOf(
        InputImage.fromBitmap(selectedImageBitmap),
        "What is in this photo?"
    )
    val response = withContext(Dispatchers.IO) {
        generativeModel.generateContent(prompt)
    }
    withContext(Dispatchers.Main) {
        OutputValue = response.text
        myTextOutput.text = OutputValue
    }
}

    This code creates an InputImage object directly from your selectedImageBitmap and includes it in the prompt array.

    1. Using ByteString (For more advanced users):

    If you want more control over the image data, you can convert your Bitmap to a ByteString before passing it to the prompt. Here’s an example:

    Kotlin
    val imageByteArray = selectedImageBitmap.toBytes("PNG") // Convert to byte array
val imageByteString = ByteString.copyFrom(imageByteArray)
val prompt = arrayOf(
    imageByteString,  // Raw image data
    "What is in this photo?"
)
// Rest of the code remains the same

    Explanation:

    • InputImage.fromBitmap(selectedImageBitmap): This creates an InputImage object specifically designed for the Google AI Client SDK, ensuring compatibility.
    • ByteString: This class represents raw byte data, allowing you to pass the image data directly if you prefer.

    Additional Notes:

    • Make sure you have the correct version of the Google AI Client SDK dependency in your build.gradle.kts file.

    Remember, these are just two options. Choose the one that best suits your needs and coding style.

    • 1

Related Questions