Bitmap generator paired with an neural network image recogniser? - neural-network

I'm new to programming, and i would be great if someone could point me in the right direction.
Hoping to create this project, but not sure whether to try and complete the programming myself or maybe pay someone to create it for me.
Hoping to create a bitmap generator that creates small (300x300px ish) images (similar to this ... RANDOM.ORG - Bitmap Generator ), which then feeds these to an neural network based image recogniser, which has been trained to pick out particular images it thinks matches an initial set of black&white bitmap images I give it to learn.
Basically, I'd like to create a program that I can feed a set of images, and it then generates images based on these, with similar tones, patterns and shapes..
Before I even start thinking about making this, do you reckon it would even work? The generator would have to be pretty quick for the image recogniser to have any hope of finding anything. And I'd have to be able to skew the percentage of black/white bytes in order to come up with the sort of images I feed it to learn.
If anyone could offer any suggests on where to start, that would be amazing!

Related

CoreML Image Detection

I want to implement an application, that is able to recognize pictures from camera input. I don't mean classification of objects, but rather detecting the exact single image from given set of images. So if I for example have an album with 500 pictures, then if I point a camera to one of them, then application will be able to tell it's filename. Most of tutorials I find about CoreML is strictly for image classification (recognizing class of object) and not about recognizing exact image name in camera. This needs to work from different angles as well, and all I can have for training the network is this album with many different pictures (single picture for single object). Can this be somehow achieved? I can't use ARKit Image Tracking, because there will be about 500 of these images, and I need to find at least a list of similar ones first with CoreML / Vision.
I am not sure, but I guess perceptual hashing might be able to help you.
It works in a way that it makes some fingerprint from the reference images, and for a given image, it extracts the fingerprints as well, and then you can find the most similar fingerprints.
in this way, even if the new image is not 100% as the image in the dataset, you still can detect it.
It is actually not very hard to implement. but if you would like, i think phash library is a good one to use.

How to detect contours of object and describe it to compare on server with ARKit

I want to detect shape and then describe it (somehow) to compare it with server data.
So the first question is, is it possible to detect shape like blob with ARKit?
To be more specific, let's describe my usecase generally.
I want to scan image by phone, get the specific shape, send it on server, compare two images on server (server image is the real one, scanned image would be very similar) and then send back some data. I am not asking about server side, the only question about server side is what should I compare - images using OpenCV, some mathematical description of both images and try to find similarity, etc.).
If the question is hard to understand, let's split it on two easy questions:
1) How to scan 2D object by iPhone and save it (trim the specific shape from its background when object is black and background white).
2) Describe scanned object for comparision with almost the same object.
ARKit has no use here.
You will probably need a lot of CoreImage (for fixing perspective distortion and binarization) and OpenCV logic.
Perhaps Vision can help you a little bit with getting ROI from the entire frame, especially if the waveform image is located in some kind of rectangle.
Perhaps you can train a custom ML model that will recognize specific waveforms or waveforms in general to use with Vision.
In any case, it is not a trivial task.

Copy face from Image

I'm a noob to this forum, but wanted to give it a try.
I'm currently learning Objective-C and Cocoa; trying to build my first iPhone app.
One thing I'm working on is allowing the user to cut his/her face from an image they have taken and paste it into another image. (The idea is cut from one image and paste into another image with a spot for a face to go.)
How can this be done? I am thinking I would allow the user to just touch and drag over their face, in the shape of a rectangle, and then allow them to copy.
Thanks for the help.
Ok, nevertheless your bit arrogant style of asking, here are some guidelines about how to start: generic obj-c/iOS development (start from hello world); UIImage class; camera API; image processing algorithms, face detection algorithms. Go on gradually and do not wish to resolve all problems at once. Write first an application that simple loads an arbitrary photo and shows it to the user. Then modify it that you can crop a specified rectangular area from the image and save it into the new file. Then write an app that switches on the camera that you can take an image and save it to the disk. Then unite what you wrote that you save only a cropped area of the captured image.
When you arrive to this point, you will know much more about software development image handling. AFTER THIS you can start looking for image processing algorithms. Start also here with something simple like a trivial blur filter or similar implemented by you. If you know already a bit of image processing, search for face detection algorithms on the net. It is even possible that you will find some ready framework that includes also these features, or at least you will understand the concepts. You can even come back here to stack overflow and ask for suggestions about a good face detection algorithms, however we still prefer if you have chosen already one and have some concrete issue with it.

compare one image in matlab with a database of images and show the most similar

I have a database of images of one person who is using his hands to show various words and phrases in sign language. The background is white and the only thing changing is the shape of the person's hands and their locations. Now in my gui in matlab, I want the user to be able to choose another image from the same person that was taken at another time doing a sign but wearing the same clothes and then the program will have to compare this against the images in the database and show the most similar. Obviously I can't do pixel by pixel comparison as the images were taken by a hand held mobile camera and slight movement has been inevitable so I should try and locate the hands in the images and compare their shapes. I have no idea how to go about this? I have to say I am new to image processing toolbox in matlab.
Your help is much appreciated
I am doing a phD in computer vision, and I can tell you that it is an unsolved problem. (even in your simple framewrok, with white background)
If you are interested, you might read some works about it ar MIT:
http://people.csail.mit.edu/rywang/handtracking/
or at Oxford:
http://www.robots.ox.ac.uk/~vgg/research/sign_language/index.html
http://www.robots.ox.ac.uk/~vgg/research/hands/index.html
I disagree with you. Such a project can achieve results quickly.
This becomes a problem as soon as the project has to deal with "real life".
Using a single camera, and a completely known background; Opencv provides a simple way to extract hand shape in a image (in about 20 lines of code). You will find plenty of source on the web (have a look at calcbackproj).
After that, what you will have to do is to play with shape, and search for characteristic points.
Begin with some simple signs (example : a circle and a V). How would you recognize one from the other?
There are thousands of papers on sign language; just read the older one to simple ideas flowing :)

Fake long exposure on iOS

I have to implement long exposure photo capabilities to an app. Since i know that this is not really possible i have to fake it. It should work like "Slow Shutter" or "Magic Shutter".
Sadly i got no clue how to achieve this. I know how to take images with the camera (through AVFoundation) but i'm stuck at merging them to fake long shutter times.
Possibly i need to manipulate and combine all the images with coregraphics but i'm not sure about this (even the how). Maybe there's a better solution to this.
I would appreciate every help i can get here,
thank you people!
You might try the plus lighter blend mode.
Well, I suppose it would be possible to average together the results of several shots. I've mucked around a bit with the core graphics stuff to resize images (averaging together adjacent pixels), but with lower res images. The algorithm I used is here -- maybe it'll give you some ideas.
There may, of course, be a better way, and some tricks for working efficiently with high-res images. Can't help you there.
Convert the images to pixel bitmaps. Align and stack the bitmaps. Then try applying various 3D convolution filters to the 3D pixel array.