ARCore Expert Needed for Enhancing Image Recognition in Android App

I am looking for an experienced ARCore developer to help diagnose and improve image recognition reliability in an Android app.

The app is being developed in Android Studio and currently uses ARCore Augmented Images.

The goal is to recognize a printed image through the phone camera and then play a video aligned on top of that printed surface.

I started by testing the official ARCore sample project.

The default sample image is recognized successfully, but when I replace it with my own custom image, recognition becomes unreliable.

For discussion purposes, you can assume the target image is something like Leonardo da Vinci's Mona Lisa.

The issue is that the image may be recognized when displayed on a computer monitor, but often fails when the same image is printed and scanned with a phone camera.

At this stage, I am not focused on final card size yet.

The final product may later use a card size around 90 mm x 50 mm, but right now the main priority is simply to achieve stable recognition of a printed image in any practical size.

What I need help with:

Diagnose why my custom printed image is not being recognized as reliably as the default ARCore sample image

Improve ARCore image recognition reliability for printed targets

Advise on the best type of images to use for Augmented Images

Identify whether the problem is related to:

insufficient keypoints/features

low grayscale contrast

repeated patterns

print quality

reflection or surface finish

camera distance/angle

incorrect reference image setup or image database preparation

Recommend a practical workflow for testing and validating target images before deployment

Current observations:

The ARCore sample image works

My custom replacement image is much less reliable

Recognition is often better on a monitor than on a printed version

Adding corner marker-like graphics did not solve the issue

This is intended for a real printed AR product, not just an on-screen demo

Ideal freelancer:

Strong experience with ARCore Augmented Images

Experience with Android Studio / Kotlin or Java

Familiar with real-world printed target testing

Able to explain clearly what is wrong and how to improve recognition performance

Deliverables:

Analysis of the current recognition issue (no using ArUco marker)

Recommendations for improving target image design and print conditions

Updated code based on the augmented_image_java sample project to improve recognition performance.

Clear best practices for achieving stable printed image recognition

If you have worked on ARCore printed image recognition before, please include relevant examples in your proposal.

Contract duration of 1 to 3 months.

Mandatory skills:

Android, Smartphone, Android App Development, iOS, Mobile App Development