OpenCV library is used by other applications for image enhancement, panorama stitching, object detection and recognition, etc. OpenCV Manager provides the best version of the OpenCV for your hardware. It also receives the latest stability and performance updates for the library.
In order to utilize the app, the latest version of OpenCV Manager (also found on Google Play) must be downloaded and installed correctly. This app has been successfully tested on Android 2.3.4 and Android 4.3.
At the moment this can considered an initial tech demo, please feel free to play around with it. The source code for this implementation can also be found in github link below!
This app compares the reference image to the images in the added list. It uses ORB feature detector with BRIEF descriptor extractors. It uses BRUTEFORCE_HAMMING method to match the descriptors. Also, based on preference, app can also check the homography of the matched keypoints with LMED (least of median squares) and remove the outliners to give a better result.
How to use:
1 - Move the vision wherever you want, take pics along the way by using ADD button.
2 - You can also take a reference picture, that you want to be recognized. When you click it, it will be shown on the right of the camera view.
3 - When you are done, and make sure you have added images and have a reference image, click the Find Match button. It will calculate the keypoints of each image and the reference image and apply the matching algorithm.
4 - After matching and comparing, it will show the best match among all matches.
5 - You can also use Homography Off/On switch to provide more accurate results.
6 - Also, you can use Images only/with matches switch to show which keypoints are matched. REMEMBER, you have to click find match again to see the effect of Homography and Matches switch.
This app is still in development, actually being used by another project but I put it on for cool demonstration!
Source code available: https://github.com/mustafaakin/image-matcher
We parallelized the algorithm using the Vector Fabrics Pareon tool. The results are shown for single, two and four application threads.
This demo should be run on dual or quad core machines.
This demo is only interesting to programmers who need to write parallelized C or C++ code.
OpenCV Manager needed . If not installed, the application will ask download.
It takes at least two faces saved so you can begin to recognize
Training Mode: Write the name of the person, focus and when it begins to appear a box locating a face press "Rec". Press Rec repeatedly to store different gestures
Find mode. Focus on one face and if recognized, its name appears. An icon will appear green, yellow or red depending on the degree of confidence in recognizing
Button "View All": See the faces stored.
+ Version 2.2: Storage moved to internal storage.
Source code: https://github.com/ayuso2013/face-recognition
Load scaled pictures
Color plane extraction - RGB, HSV, HSL, XYZ, YCrCb
Binary convolution functions like Dilate, Erode, etc
Save post-processed pictures
Camera Classifier é um aplicativo que permite realizar a classificação de imagens utilizando o descritor BIC, com classificador 3NN com a biblioteca OpenCV. Seu código fonte é aberto e disponível em https://github.com/nihey/CameraClassifier
Simple edge detection should be fluent for everybody but face detection need more processing power.
Use Neon for colorspace transformation
Face detection optimization is done thanks to Bill Mc Cord