I am currently evaluating capabilities of IBM Watson Visual Recognition service to recognize faces. So that System should identify the each person that we have trained. Individuals may come with different clothes, and other possible variations. But system should identify each individual by looking at each face.
As per IBM, IBM visual recognition do not support face recognition but only face detection.
Face Recognition: Visual Recognition is capable of face detection
(detecting the presence of faces) not face recognition (identifying
individuals).
Can we use the custom classifiers by adding different types of images for each individuals?
What is the significant pre/post-work from the developer to get at least 90% accuracy ?
Matt Hill posted a great reply to this similar question on dW Answers. Here's what he had to say:
It is possible to train a custom classifier to try to identify people's faces. It might help to use the face detection service as a preprocessor to give you bounding boxes around faces, and use them to crop the images submitted for custom classification. However, the VR custom learning engine is not optimized for face identification, and I would not expect the results to be as accurate as a system that is designed specifically for face recognition.
The issue is that human faces are typically very similar to each other in with respect to the wide set of features that were trained in learning the basis of the system, which needed a very broad exposure to many types of scenes and objects.
Related
I am working on a crowd controlled soundsystem for a music festival. Music would be controlled by individuals and the crowd as a whole, more or less 500 people.
While searching for crowd tracking techniques, I stumbled upon this one http://www.mikelrodriguez.com/crowd-analysis/#density; Matlab code and dataset are enclosed. Are you aware of similar techniques, maybe simpler, based eg on blob detection? Do you have an idea about how well this one would perform in a real-time scenario? Is there a known way to do this with eg OpenCV?
One of my former colleagues implemented something similar (controlling a few motors according to crowd movement) using optical flow. You can analyze the frames of video from a camera, calculate optical flow between frames, and use the values to estimate the crowd movement.
OpenCV has support to perform the above tasks, and comes with good code samples. A desktop should be able to do this in real-time (you might have to tweak with image resolution).
I am not exactly sure how to interface between a C++ program and a sound system. Pure Data (PD) is an alternative, but it might not have much support for motion analysis.
I am starting an AR project for a client which involves using AR in order to show information about certain objects. In this project, for example, the user would point the camera at a car. Depending on which part of the car the user is looking at (headlights, windshield) a button would appear. When the user presses that button, an information window would appear on screen, giving the user more information about that certain car part.
The client doesn't wish to place physical markers on the car (QR code / patterns), and so the car parts would have to be detected another way.
I have developed AR apps before, but based on user location and generated markers in the sky. I feel this system wouldn't be entirely relevant for the client's request.
Would anybody be able to point me in the right direction (iOS library) for this sort of project, and whether or not it would be entirely feasible.
Thanks for the input,
Andy.
What you need is a model-based tracker/6DOF object tracker. As you want to track a car, it will certainly be featureless (or you will only get sparse features), so you should look at textureless non planar 3D (object) tracking solutions.
It's pretty much state of the art right now (lot of research, few products/SDK), but using library like OpenCV and with the appropriate literature (see below) you should be able to develop one. You can look at an open-source solution like the ViSP library which has a module for model based tracker but not an official iOS port. for commercial libraries, closest will be AR libraries supporting SLAM or "3D object tracking".
In term of techniques, you have different way to handle this problem, some pointers:
You can use a model-based tracker relying on edge detection + initial CAD model of the object: 3D Textureless Object Detection and Tracking: An Edge-based Approach or Harald Wuest, Folker Wientapper, Didier Stricker Adaptable Model-based Tracking Using Analysis-by-Synthesis Techniques
The 12th International Computer Analysis of Images and Patterns (CAIP), 27-29th August 2007, Vienna, Austria.
You can use a model-based tracker relying on edge detection + (trained) template images
You can use some SLAM techniques combined with a model based tracker.
M. Tamaazousti, V. Gay-Bellile, S. Naudet Collette, S. Bourgeois, M. Dhome Real-Time Accurate Localization in a Partially Known Environment:Application to Augmented Reality on textureless 3D Objects. TrakMark 2011, Basel, Switzerland 26-29/10/2011
if your system will only run indoor, you can look at some RGBD tracker
S. Hinterstoisser, V. Lepetit, S. Ilic, S. Holzer, G. Bradski, K. Konolige, N. Navab
Model Based Training, Detection and Pose Estimation of Texture-Less 3D Objects in Heavily Cluttered Scenes Asian Conference on Computer Vision (ACCV), Korea, Daejeon, November 2012
(access to the software)
It seems you are heading for an interesting topic. However, my concern is the accuracy of what you are trying to do. Location-based AR would be a starting point for your research work. Still the granularity would be less to your problem domain. Since you have worked on the location-based AR application, you might have noticed the accuracy that you can expect would be maximum upto 3 meters. Therefore, that level of accuracy cannot address your problem domain in an advanced way.
However, I have seen prototypes that addresses your problem domain. One good example would be the BMW Augmented Reality Manual. Check this link http://www.youtube.com/watch?v=P9KPJlA5yds
Hence, I never came across a proper Augmented Reality library for iOS or even Android which can address your problem domain in the marker-less AR context.
The information above is only for your knowledge, but not to discourage you in any way.
I would like to know what is the best approach to begin a project to perform graphical recognition of people. In other words, the computer will parse an image file and through a heuristic figure out if it sees the shape of a person.
Any API's or open sources available, is this too ahead of the times?
Thanks
Are you searching for face detection or people detection?
If face-detection:
OpenCV comes with samples for face detection. And OpenCV 2.4-beta has samples for face recognition also. Check here : http://github.com/Itseez/opencv/tree/master/samples/cpp
If people-detection:
OpenCV comes with a sample for people-detection using HOG descriptors. Link
This is the result i obtained with above code:
Hey guys, Am wondering if anybody can help me with a starting point for the design of a Neural Network system that can recognize visual patterns, e.g. checked, and strippes. I have knowledge of the theory, but little practical knowledge. And net searches are give me an information overload. Can anybody recommend a good book or tutorial that is more focus on the practical side.
Thank you!
Are you only trying to recognize patterns such as checkerboards and stripes? Do you have to use a neural network system?
Basically, you want to define a bunch of simple features on the board and use them as input to the learning system. It can often be easier to define a lot of binary features and feed them into a single-layer network (what can become essentially linear regression).
Look at how neural networks were used for learning to play backgammon (http://www.research.ibm.com/massive/tdl.html), as this will help give you a sense of the types of features that make learning with a neural network work well.
As suggested above, you probably want to reduce your image a set of features. A corner detector (perhaps the Harris method) could be used to determine features in the checkerboard pattern. Likewise, an edge detector (perhaps Canny) could be used in the stripes case. As mentioned above, the Hough transform is a good line detection method.
MATLAB's image processing toolbox contains these methods, so you might try those for rapid prototyping. OpenCV is an open-source computer vision library that also provides these tools (and many others).
For a while I've been attempting to simulate flowing water with algorithms I've scavenged from "Real-Time Fluid Dynamics for Games". The trouble is I don't seem to get out water-like behavior with those algorithms.
Myself I guess I'm doing something wrong and that those algorithms aren't all suitable for water-like fluids.
What am I doing wrong with these algorithms? Are these algorithms correct at all?
I have the associated project in bitbucket repository. (requires gletools and newest pyglet to run)
Voxel-based solutions are fine for simulating liquids, and are frequently used in film.
Ron Fedkiw's website gives some academic examples - all of the ones there are based on a grid. That code underpins many of the simulations used by Pixar and ILM.
A good source is also Robert Bridson's Fluid Simulation course notes from SIGGRAPH and his website. He has a book "Fluid Simulation for Computer Graphics" that goes through developing a liquid simulator in detail.
The most specific answer I can give to your question is that Stam's real-time fluids for games is focused on smoke, ie. where there isn't a boundary between the fluid (water), and an external air region. Basically smoke and liquids use the same underlying mechanism, but for liquid you also need to track the position of the liquid surface, and apply appropriate boundary conditions on the surface.
Cem Yuksel presented a fantastic talk about his Wave Particles at SIGGRAPH 2007. They give a very realistic effect for quite a low cost. He was even able to simulate interaction with rigid bodies like boxes and boats. Another interesting aspect is that the boat motion isn't scripted, it's simulated via the propeller's interaction with the fluid.
(source: cemyuksel.com)
At the conference he said he was planning to release the source code, but I haven't seen anything yet. His website contains the full paper and the videos he showed at the conference.
Edit: Just saw your comment about wanting to simulate flowing liquids rather than rippling pools. This wouldn't be suitable for that, but I'll leave it here in case someone else finds it useful.
What type of water are you trying to simulate? Pools of water that ripple, or flowing liquids?
I don't think I've ever seen flowing water ever, except in rendered movies. Rippling water is fairly easy to do, this site usually crops up in this type of question.
Yeah, this type of voxel based solution only really work if your liquid is confined to very discrete and static boundaries.
For simulating flowing liquid, do some investigation into particles. Quite alot of progress has been made recently accelerating them on the GPU, and you can get some stunning results.
Take a look at, http://nzone.com/object/nzone_cascades_home.html as a great example of what can be achieved.