NeuraLens

This project was build during "Hackaton Google : Solve for Healthcare & Life Sciences with Gemma" the 07/08/2025.

Fine Tuning setup

create a .env with HF_TOKEN

conda create -n medgemma python=3.10.17
pip install -r requirements.txt

Then use the notebook from notebooks/finetune.ipynb

The Fine tuned model is also direcly available on huggingface https://huggingface.co/MasterTrtle/Merge-medgemma-4b-it-lora-tissue-classifier

Downloading directly the app:

https://drive.google.com/file/d/1IfqtY7GVd-CdcJSUv8jDuaYx3vD_pULk/view?usp=sharing

App Building

Building and Debugging

This project uses Gradle for building. You can use Android Studio or the command line to build and debug the app.

Prerequisites

• Java Development Kit (JDK) 17 or higher
• Android Studio (latest version recommended) or Android SDK command-line tools

Building the APK

To build the debug APK from the command line, run the following command in the root directory of the project:

./gradlew assembleDebug

The generated APK will be located in app/build/outputs/apk/debug/app-debug.apk.

Installing the APK

To install the APK on a connected Android device or emulator, use the Android Debug Bridge (adb):

adb install app/build/outputs/apk/debug/app-debug.apk

Debugging

You can view the application logs for debugging purposes using adb logcat:

adb logcat

To filter the logs for this specific application, you can use a command like this:

adb logcat com.example.ollamacameraapp:V *:S

Configuration

Before running the application, you need to add your Vertex AI access token in app/src/main/java/com/example/ollamacameraapp/MainActivity.kt.

Replace "YOUR_ACCESS_TOKEN" with your actual access token.