Closed. This question needs to be more focused. It is not currently accepting answers.
Want to improve this question? Update the question so it focuses on one problem only by editing this post.
Closed 6 years ago.
Improve this question
I'm going to be more specific about the situation:
I've captured a screenshot from the game DotA. The information I want to get is what objects eg. heroes (also its name, hp, ...), creeps (also which side), towers, etc. is visible in the image and where they are. A problem come from the fact that in DotA 2 many of these object can be viewed from many perspective, so let's reduce the problem and assume that every object have only one orientation. How might this problem be solved quickly enough, that it can recognise all objects in real time at about 30fps? Any help or suggestions is welcome.
I think that you have the good flags: CNN for image segmentation. So my point is that for so many different objects from different points of view and scale (because I guess that you can zoom in/out on your heroes/objects), the easiest way (but the heaviest in term of computation) is to build one CNN for each type of object.
But images would help a lot to get a better understanding of the problem.
Related
Closed. This question needs to be more focused. It is not currently accepting answers.
Want to improve this question? Update the question so it focuses on one problem only by editing this post.
Closed 6 years ago.
Improve this question
I want to build an app that locate and track people by iPhone using Gyro+Accelerometer. No need for GPS here.
How should I approach this problem?
Unfortunately it's not feasible to track device position based on accelerometer and gyro.
In order to calculate position from accelerometer data double integration needs to be applied, integration amplifies noise and turns it to drift, so even small measurement error would create huge position drift. Similar problem appears for gyro as well.
You can find more details here:
http://www.youtube.com/watch?v=C7JQ7Rpwn2k&t=23m20s
You don't!
The values retrieved from the gyroscope and accelerometer are "relative" to the device. They have no absolute meaning which you would need to retrieve some kind of location.
You can theoretically measure / calculate what way a device has taken but you do not know if the user took that way in Germany, China or the USA. You know he went right, then left, then 200m straight - but that is not of any help if you do not know from where he originated.
That being said, if you do have the initial position you can theoretically calculate the new position based on the measured values. But that calculation is probably far to error prone and far to inexact. If you try to measure the values over the course of a few minutes or even hours you will probably get an measurement that is many meters or even kilometers off.
Closed. This question needs to be more focused. It is not currently accepting answers.
Want to improve this question? Update the question so it focuses on one problem only by editing this post.
Closed 7 years ago.
Improve this question
I'm stuck on a problem and needed some help or guide for a possible solution.
Basically in my application there will be a map with several zones.
The user can select any of these areas, at that time this area is filled with a color.
Imagine a map like this one, so i need to be able to change the color of only one country.
Something like what happens in the books of paintings (https://itunes.apple.com/pt/app/colorfly-best-coloring-book/id1020187921?mt=8), or Paint Bucket command in the Photoshop .
Any idea how to get something like this on iOS ?
Thanks in advance
The paint bucket technique you're looking for is a set of graphics algorithms usually called "flood fill". There are different approaches to the implementation depending on the circumstances and performance needs. (There is more at that wikipedia link.)
I have no experience with it, but here is a library from GitHub that purports to implement this for iOS given a UIImage object: https://github.com/Chintan-Dave/UIImageScanlineFloodfill
Re: your question about doing this without user touch: yes, you'll want to keep a map of countries to (x,y) points so you can re-flood countries when required. That said, the intricacies of the county borders might make an algorithmic fill inexact without more careful normalization of the original source. If your overall map only contains a small set of possible states, there are other ways of achieving this goal, like keeping a complete set of possible images (created in ie Photoshop) and switching them out, or keeping a set of per-country "overlay" images that you swap in as needed. (But if the flood fill is accurate on that source image, and performant for your needs, then great.)
Closed. This question needs to be more focused. It is not currently accepting answers.
Want to improve this question? Update the question so it focuses on one problem only by editing this post.
Closed 7 years ago.
Improve this question
I want to dynamically create realistic clouds including their movement and multiplication (density). I found this link which seem to explain more than I can understand. I cannot find a way to replicate the clouds and there is practically no other tutorial online. Does anyone know how can I achieve this result? Thank you in advance!
use scene kit you can make a dynamic particle effect:
https://developer.apple.com/library/prerelease/ios/samplecode/SceneKitVehicle/Introduction/Intro.html
in this example you can find, after the vehicles code, the smoke animation. You can gat all the stuff you need from there, like smoke.scnp, smoke.png.
You need to change some row of code like the smoke color, the light color and the camera position.
But with that i guess you can have one of the best result possible, as real time cloud simulation, in iOS.
(other solution would be using openGL but trust me this is a ton easier)
maybe this guide can also be useful to understand the basis (you just need them to achieve your clouds) :
http://www.raywenderlich.com/83748/beginning-scene-kit-tutorial
Closed. This question needs to be more focused. It is not currently accepting answers.
Want to improve this question? Update the question so it focuses on one problem only by editing this post.
Closed 3 years ago.
Improve this question
I have coded a program which dynamically inserts nodes and edges, overlaying the graph onto an image. My aim is to be able to automatically add the edge weights based on pixel distances from one node to the next (edge length). Is this possible? If so could you guide me in the right direction. Thank you.
JUNG doesn't exactly have a native notion of edge weights (or any other edge- or vertex-related metadata). What it has instead is a convention for how to tell algorithms that need such metadata how to access it. For more information, see the "User Data" section here: https://sourceforge.net/apps/trac/jung/wiki/JUNGManual
Closed. This question needs to be more focused. It is not currently accepting answers.
Want to improve this question? Update the question so it focuses on one problem only by editing this post.
Closed 7 years ago.
Improve this question
I would like to convert a powerpoint presentation into a series of images. One per slide specifically so they can be uploaded as an image gallery to a blog. Does anyone know of any libraries that can convert a .ppt into images. Any language is fine as long as it can run on a *nix server, so no C# or .Net dependent libraries.
Even if one existed, and I would guess one does, it can't address animations in a meaningful way. I see more and more ppt presentations making good use of animations to get their points across. Many of these animations will overlap one another. How could such a slide be turned into a single image? How would you prioritize which animated image segments should overlap the others? You may want to keep an eye on this thread: Converting ppt to png using Apache poi
This might get you a solution via php How to work with powerpoint in php?
This might work with python: https://stackoverflow.com/q/4995877/657003
This is a good discussion of the problem you face: http://www.joelonsoftware.com/items/2008/02/19.html