openftc / opencv-repackaged Goto Github PK

We are trying to use EasyOpenCV on a Moto G5 Plus running Android 7.0 . The libOpenCvNative.so file was copied using the ADB PUSH command to the /sdcard/FIRST folder. On staring the Robot Controller app we are getting the following error - " libOpenCvNative.so is corrupted - libOpenCvNative.so is present in the FIRST on the internal storage. However, the MD5 checksum, does not match what is expected. Any OpenCV-enabled OpModes will likely crash. Delete and re-download the file."

Running the following command using adb shell -
md5sum libOpenCvNative.so
returns
07a377983266cb64bd649bb233452970 libOpenCvNative.so

Any hints to what may be causing this ?

The AndroidManifest.xml file has old (4.1.0) opencv version info.
This can be corrected by removing the following lines from the manifest:

android:versionCode="4100"
android:versionName="4.1.0">

and adding the definitions to the build.gradle before the android section:

def openCVersionName = "4.5.3"
def openCVersionCode = ((4 * 100 + 5) * 100 + 3) * 10 + 0

and the reference on those in the android.defaultConfig section:

versionCode openCVersionCode
versionName openCVersionName

The compileOnly reference to RobotCore should also be updated to 7.0.

Finally, it may be possible to use the annotation org.firstinspires.ftc.ftccommon.external.OnCreate on the loadNativeLibOnStartRobot method as opposed to the opModeRegistrar. Not sure it is a huge difference, but may save some attempted reloads.

Our team is interested in using code from opencv-contrib modules, specifically org.opencv.aruco. Is there anyway to do this while using EasyOpenCV and OpenCV-Reapackaged?

I am fairly new to android and FTC. Last year I successfully used EasyOpenCV (looks like it was v 1.5). This year I plan on using EasyOpenCV again, and started by using your latest version. Initially, I'm getting an error on the import I used last year (import org.openftc.easyopencv.OpenCvCamera). I'm guessing the package naming changed with the new version. Last year there was some sample code in an "examples" folder that was a great help. Any chance that could be added for the updated version?

Also, from the instructions, I wasn't sure which libOpenCvAndroid453.so to use (the one in arm64-v8a or the one in armeabi-v7a). In addition, last year there were instructions to remove the arm64-v8a from the build.common.gradle file. Is that still an issue?

When I start the Robot Controller, I get
Could not load OpenCV native library because app is running in 64-bit mode. Please remove the arm64-v8a entries from build.common.gradle
When I remove the arm64-v8a entries it works fine. Just wondering if there is a 64 bit version of this library somewhere?

In readme.md, it says that libOpenCvNative.so should be put in android internal storage. But it does not make sense since the internal storage of an app is hidden for the users. The only thing we can access from the file system is an app's external storage.

Accured: putting libOpenCvNative.so in internal storage

Expected: putting libOpenCvNative.so in external storage

ref: https://stackoverflow.com/questions/5092591/what-are-the-differences-among-internal-storage-external-storage-sd-card-and-r

Md5 signatures are found on libs in /doc, but there are no md5.txt files in this project. They'd be useful to check for file corruption.

Right now you instruct users to copy libOpenCvNative.so to a specific directory on the device and then you load the shared object from that specific path via System.load. This adds an extra step to the setup process, and I was wondering if there was any reason you don't place the shared object under src/main/jniLibs/{ABI TRIPLE}/ where it will be automatically included in the AAR?

If there's no reason that this isn't being done then I'll go ahead and make a PR that packages the shared objects in this manner.

The control hub has a read-only storage when connected to a computer via TypeC-USB, so files cannot be added to the FIRST folder.