Assuming i have a working facial recognition algorithm working for the iphone (comparison images would be stored on my machine). How can i expand this to compare image 'A' against images stored on a remote server?
Could someone give me an abstract definition? (I could download and temporarily store all images on my iphone, and then compare them against image 'A' however if i have hundreds of photos then it will take too long to process and be useless..).
Think like google goggles does - upload your image to the server and let the server crunch it. You can optimize the upload by sending a pre-processed part of the image first.
What is the input to your comparison algorithm, is it the subwindow containing the face detection? You could, just like peterept said, upload THAT smaller subwindow containing the face to the server and let the server do the work. If your input to the comparison algorithm is a set of features extracted from the face (statistics, etc.), then if it doesn't take too much CPU power to process the subwindow, you could extract those features on the phone and send them to the server for processing.
The main idea here is that you let the remote server do all the crunching, even if it is hundreds of photos (will still be faster than the phone, especially if you multithread). Then the trade-off is between sending the subwindow image or the feature set extracted from that subwindow, whichever's smaller to send.
Related
I looking for the best way to upload an image from mobile phone to my server. I am currently using html5 to open the camera and take the picture, then I convert the file into a base64 string, then I send to the server, then save it in MongoDB.
I am expecting around 1000 to 1500 user request per day ( upload image ) , so I have the following question :
Is it a good way to do it?
Should I compress the base64, if yes how?
Should use a specific server to handle this task?
My backend is node express and the front end is ReactJS.
Thanks
It all depends on your situation. Reading and writing images from a cdn via i.e. streams is usually faster than reading and writing binary representations of images i.e. base64 from a database. However, your speed if reading from a cdn will obviously be effected by what service you use. Today, companies like Amazon can offer storage to a very cheap price so if you are not building a hobby app for like a student project you can usually afford it. Storing binary representation of images actually end up a little bit bigger in size than storing the image itself. You don't compress the base64, you compress the image before converting it. However, if you can't afford a storage account and if you know your users won't upload that many images it is usually enough to store binary representations of the images in a database. Mongo Atlas, for example, offers 512 mb for free on their database clusters. Dividing tasks of your app such as database requests and cdn services from your main application is usually a good choice if possible. This way you will divide the cpu, memory, etc. of your hardware and it will lead to faster reading and writing tasks for the user.
There are a lot of different modules for doing this in node. JIMP is a pretty nice one with loads of built in functions like resizing images and converting them to binary, either as Buffer or base64.
What's the best way to store a very large image for an iOS app? I want an app to be able to view images that might be hundreds of megabytes, perhaps as much as a gigabyte as jpeg. I need to be able to store the image and retrieve selected areas for display.
Currently the images are cut into 512x512 pixel tiles and stored as jpeg files in a directory tree with tens of thousands of tiles (actually an image pyramid including downsamples).
Ignoring the question of displaying the image, I'm interested in the most efficient, manageable way to store this data on the device: files, like they currently are, in an sqlite database or something else?
Second part to the question. Is there a limit to the amount of data an app can store, or can an app keep importing data up to the storage limit of the device. I'm asking here about data that an app imports after it's installed.
The solution to this is to pre tile the enormous image so the tiles can be quickly retrieved from the file system on an as needed basis. One problem with very large images is that most solutions require the whole image to be rendered into a context, consuming vast amounts of memory. On a system like iOS, where memory is limited, the way to solve this is to use a library like libjeg or libjpegturbo to render an image a line at a time, then save the pixels into a raw file. The downside to doing this directly is that when you need one tile, you need to jump all over the file system finding each row of a tile. Thus a better solution is to not only incrementally scan, but incrementally tile too. You can use mmap to map the file into just the area you need, so you can really minimize memory consumption. That said, you can thrash the Unified Buffer Queue on iOS so badly the app crashes, or even the whole system!
If you are curious about how to implement the above solution, there is a freely available project on github - PhotoScrollerNetwork - that does all the above.
A sample from Apple: PhotoScroller
What about splitting into parts. Then it can be gathered by your application if needed
I need help with downloading from webserver...
What i currently do is get XML file from web servers that contains image locations, parse XML, download each image, store image on iphone, and store image name to sql database.
This takes lots of time because there is large amount of images to be downloaded and i am downloading one by one.
My app update just got rejected because reviewer decieded that downloading is too long... What is funny, last two updates passed without problems..
I was thinking about zipping those images on server and sending zip file to iphone, unzipping it there, or packing images together with binary and sending it to apple.
Any advice on how to make download faster, would be appreciated. Thanks.
BTW, zip won't help with images. They are already compressed, so it will just add overhead. Make sure your images are not any larger than you need for display and I'd do what Mario suggested above and download them in multiple async calls (at least make the one big call asynchronous.)
A key principle of UI design is to display partial results (unless they are invalid or misleading) so that the user understands that progress is being made.
If you really need all the images to make it valid, you can download a few and display them grayed out (alpha = 0.4) or something so that it's clear that this is a partial result, but that progress is being made. The reviewer probably felt that it was taking too long to startup.
Do you change those images often? Or only once per release if at all? If they change with each release only I'd package them. If they're almost never changed, go with the one huge download (so people don't have to redownload when updating) and if they're change often, download them file by file but try to do 2-3 files at once using asynchronous download (if supported).
1) I would use something like an NSOperationQueue to download around three images at a time in the background. Much more than that and the UI starts getting choppy.
2) Also display some kind of loading indicator while this is going on.
3) What format are your images in? If you are transferring over the network you should use JPG, and consider setting the quality level to something smaller (say 6 even 5). To offset the loss of quality you could send down larger images, even with the larger number of pixels you can easily be better off with a lower quality compression.
4) If you have to use PNG to preserve transparency, consider using PNGCrush on the images before sending. As noted, zip will do pretty much nothing.
One way to speed up download of those images is to put them on a CDN. Some CDNs, like Limelight have special network optimizations for sending data to mobile devices. They also just do a better job of routing content, and have higher capacity for transmitting content. What's nice about this approach is that you might not have to change your app. However CDNs can be pricy.
Likely, your images are just way too large. You said you're worried about the 20MB app limit, but I think at that point, your images are just way too large for the phone.
Rather than zipping the files, I'm pretty sure you need to downsample the size of the images. Not only that, but you should only download the ones that you need, when you need them.
If you still want to have bulk downloads, why not have it as a side option rather than the default implementation?
I don't know which is the most efficient way of organizing images downloaded from server. I will be downloading around 200 images on to my iPhone on request for download. Which is the most efficient way of organizing ? just dropping it as a file on the phone's memory or having it in sqlite (via coredata) after download ? which one is most efficient and easy to handle ? which access is faster ?
The rule of thumb is to put them (or any bigger binary data) onto disk directly, and if the whole app organizes its data with a database / CoreData, then put the paths of the images in there.
AFAIK , Iphone has minimum 8GB of memory. That will be enough for images. Also It depends upon the frequency of image downloads. If you download 200images daily then u need some application that will push it in your sqlite db. Advantage of this will be your all image files will be inside a single db. No scattered images. But if you want to store only 200 images then i would recommend it store on your phone memory with some image managing tool like ACDSEE in windows, that will help you viewing images in slide show or what ever manner you want.
I have a script that generates wave files, based on user input.
I want to be able to stream those wave files online(not necessarily as wave files, they can be converted on the fly to mp3 or whatever). Preferably through a embedded flash streamer, but a html5 version would be good too.
The files are generally small, around 5 seconds long, and I'd like then to be stream multiple files in one session.
Does anyone know how I should go about implementing this?
With such short audio clips I would not bother with a 'real' streaming technology, but just serve them up via HTTP as static files as quickly as the network connection will allow. A quick look at my iTunes library indicates that a 5s 128kpbs 44kHz stereo file is between 120-250KB. Almost small. If you are talking about 32kbps mono, then maybe the sizes will be a mere 15-30KB.
Encoding on-the-fly may result in undesirable issues, like scaling (CPU load from all those encoding jobs, some of which will be duplicate), latency (setting up the encoding, the actual encoding), and you won't know the end file size which can cause problems. So, setting up a caching system may make more sense.
I use wpaudioplayer to stream MP3s from my website (Example). It was originally made as a wordpress plugin but can be used as a standalone javascript.
I believe that it can play wave files as well as MP3s. If you do end up converting them before serving them I would suggest that you would