DogeCV

An easy to use computer vision library used for FTC Games to detect game objects. Based on EasyOpenCV and OpenCV.

Project Status:

In Development

DISCLAIMER

THIS REPO IS STILL UNDER CONTINUOUS DEVELOPMENT. WE WILL BE ADDING FURTHER DOCUMENTATION, BUG FIXES, AND NEW FEATURES

To run this library, you need an SDK version of at least 5.1

Unforunately, this library will not work on the ZTE phones, as OpenCV does not work on KitKat.

Videos

Wizards.exe have been amazing with their DogeCV coverage, so please check them out and give your support, the videos are easy to understand and well made, great for peopling who want to learn DogeCV.

• Wizards.exe Tutorial

Active Development Team

• Aditya Mangalampalli FTC 9614 Hyperion @Alpheron#2162
• Alex Carter FTC 7195 Mechanical Memes @VictoryForPhil#0001
• Arnav Komaragiri FTC 8719 Quantum Leap @-----#9037
• Ben Loan FTC 3795 JagWired @Ben6501#0094
• Daniel Goz FTC 7026 JDroids @dansman805#5805
• Frank Portman FTC 8581 Aedificatores @Something Disposable#9622

Active Contributors

• Adhit Siripurapu FTC 9614 Hyperion (Testing) @Prickles#0203
• Ishaan Oberoi FTC 9794 Wizards.exe (Publicity) @RollerCoaster45#0637
• Nathaniel Lesser FTC 12897 Newton's Law of Mass' (Logos)
• Sarthak Bhatnagar FTC 9794 Wizards.exe (Odometry & Tutorials) @s.bhatnag#0906
• OpenFTC Team for EasyOpenCV which this library is based on

Doge Alumni

• Levi Gershon FTC 12897 Newton's Law of Mass @LegoF4#2372

Other Past Contributors

• Robert Iridon and Prodaniuc Pavel FTC 15994 CSH (Webcam testing)
• Nathaniel Lesser FTC 12897 Newton's Law of Mass' (Banner art)
• Karter FTC 5975 Cybots (Brainstorming for Jewel Detector)
• Derek FTC 5484 Enderbots (EnderCV Classes)
• Owen Gonzalez (Testing)
• Kwon Paradigm Break (Testing)

Known Issues

(These issues are referring to the 2019-2020 FTC Year Detectors, and not the Relic Recovery ones as those are no longer supported)

• RelicRecovery detectors not ported
• RoverRuckus detectors not ported

Planned Features / TODO

• Expand Wiki
• Add in distance scoring between samples to increase accuracy
• General Code Clean up
• Basic angle/positioning of elements
• Port Old detectors to new systems
• Better Safety checks
• New Frame Input System
• ML detectors

Install (Credit to EasyOpenCV)

1. Pull up Android Studio, with the FTC application SDK open
2. Go to the root build.gradle
3. To the repositories section, add the lines

allprojects {
  repositories {
    maven { url 'https://jitpack.io' } // this line!
  }
}

3. Add the line implementation 'com.github.dogecv:dogecv:2020.4-alpha' to TeamCode's build.release.gradle, inside the dependencies block
4. Press the Sync Now button that should appear in the top right
5. Because EasyOpenCv depends on OpenCV-Repackaged, you will also need to copy libOpenCvNative.so from the /doc folder of that repo into the FIRST folder on the internal storage of the Robot Controller.

Detectors Status

• Stone Detector - Implemented. Needs improvement.
• Skystone Detector - Implemented. Not Reliable, Under rewrite.

FAQ

• Can I still use classic OpenCV? Yes! We want teams to use DogeCV to learn about vision and start to create their own vision systems. That's why OpenCV is open in all layers of DogeCV, and we keep the classic OpenCvPipepline introduced in the EasyOpenCV library.
• Can I use a webcam? Once more, yes! See the DogeQuickStart and the wiki for an explanation of how to this.

Changelogs

2020.1-alpha:

• Changed to EasyOpenCV from EnderCV (thanks OpenFTC team!)
• Added StoneDetector
• Added SkystoneDetector
• Removed legacy detectors (may be added back later)
• Added DigitalCamera class for localization

2019.1:

• Complete re-work of the Dogeforia system. It is now integrated within the OpenCVPipeline class, and is much simpler.
• Implemented image cropping, see CroppingExample
• Fixed miss-rotation in Vuforia display. The image should now be correctly oriented on the RC display
• Added VuMarkExample OpMode to demonstrate the basics of the new system
• Added WebcamGoldExample OpMode to demonstrate how to use a webcam with a detector under the new system
• Added `GoldAndVuMarks
• Added CroppingExample OpMode, intended to ease image cropping and to showcase functionality
• Added GoldExample OpMode
• Added method getYPosition() to GoldAlignDetector
• Added VuMarkDetector to provide a ready-made detector for the VuMarks within DogeCV. I will have it highlight VuMarks at a later date
• Added enum VuMark to DogeCV to make dealing with VuMarks easier in code
• Added enum 'CameraModetoDogeCVto make setting the camera betweenFRONT, BACK, and WEBCAM` easier
• Added wiki entries for VuMarks and Webcams
• Reworked code comments
• Removed camera index as a constructor argument in OpenCVPipeline. This has been replaced by DogeCV.CameraMode
• Updated page banner - thanks Nathan!
• Updated this README

2018.2.1 HOTFIX:

• Removed AsyncFilterRunner import from Sampling Detector #12
• Fixed Dogeforia crash on stop #10

2018.2:

• Added HoughSilverDetector: intended for slighter slower but more precise silver mineral classification. Works nicely.
• Added JavaDocs and in-depths comments on most detectors, scorers, and filters
• Changed DogeCVScorer input from MatOfPoint to Mat in order to facilitate more diverse scoring methods
• Detectors now have a seperate displayMat as opposed to a workingMat; one is for detection work, the other is for displaying the results
• Added SilverExample OpMode for demonstrating silver detector use
• Added HoughSilverExample OpMode for demonstrating Hough transform-based silver detector use
• Gold Align Detector now has setAlignSettings(int offset, int width) function to set parameters
• Updated SamplingOrderDetector to choose top two silver minerals instead of all
• Ported GenericDetector to using DogeCVDetector
• Ported BlankDetector to using DogeCVDetector
• Added updateSettings to LeviColorFilter
• Corrected typos within this README
• Added an experimental WHITE option to LeviColorFilter
• Cleaned up detectors and filters

2018.1:

• Added HSVRangeFilter that uses classic lower and upper HSV ranges
• HSVColorFilter now divides the range by 2 on each side of the perfect color value. (Now range acts as expected) (Issue #9)
• Added SilverDetector (Same as Gold but for silver)
• Moved SamplingOrderDetector to use HSVRangeFilter for silver
• Tuned Sampling (Still not competition ready)
• Cleaned up Dogeforia handling per suggestions of "@BillTheCat123 | Mentor | 3763" on FTC Discord
• Fixed Gradle import issues of :FTCRobotController with Dogeforia class. (Issue #8)
• Cleaned up Gradle Build files to lessen import errors
• Added common Gradle Build errors to wiki
• DogeCVDetector now handles printing the detector setting to the screen

2018.0:

• New Versioning System
• New Scoring API
• New DogeCVDetector Class
• Vuforia Support
• Gold Align, Mineral Order, and Gold Detectors
• General Code Cleanup
• Cleaner Params
• Moved Downscaling to DogeCVDetector Class
• Updated to EnderCV 2.0 (Modfied)

1.1.1 HOTFIX:

• Fixed Jewel Detector Blue Filter

1.1:

• New Color Filter API
• New Generic Detector
• Fixed Jewel Debug Scores
• Fixed Imports for DogeLogger inside Cryptobox Detector
• Ported all detectors to Color Filter API
• Added Yellow to LeviColorFilter
• Added HSV color filter
• New Relic/Generic Example

1.0:

• New Cryptobox Detector
• YouTube Tutorials
• Per-Detector Documentation
• Wiki Start
• Added perfectRatio tuning for Jewels
• Optimization
• Removed Multiple Mat returning

0.5:

• Fixed rotated preview on portrait mode.
• Detectors return an array of images. You can cycle through them by tapping on the preview screen

Contact

Please feel free to message us on Discord with the usernames that we have listed above. You can also usually spot us on the FTC Discord and feel free to ask us if you have any questions.