Autopano Giga 2.6 review – big prints
Kolor Autopano Giga 2.6 review
Image stitching and large panoramic prints
...Get our Newsletter for new articles/reviews and why not subscribe to Keith's YouTube Channel
...Keith's book about how to use tilt/shift lenses is now available.
Our site contains affiliate links - these help support the site. See our Advertising policies for more
Keith looks at V2.6 of the ‘Autopano Giga‘ image stitching software from Kolor.
This software actually covers a wide variety of potential usage options for panoramic images, but here, Keith is concentrating largely on producing high quality images for making large prints, both from a landscape and architectural photographer’s point of view.
Large prints in this instance are only limited by the capabilities of our Canon iPF8300 printer at 44″ width and (up to) 15 metres long…
This review is based on using the Apple Mac version of the software, but the Windows and Linux versions are similar.
There is a full list of functional specifications for the software later in this review, including the Autopano Pro version of the software.
There is a downloadable demo available.
Feb. 2013 – APG 3 review . Wider range of projection geometries and noticeably faster (if you are new to APG, I’d suggest to read this review first)
For many years I’ve taken multiple shots at locations, often with the thought that I could stitch them together at some time and potentially produce an image that matched more of my feel for how the scene looked to me.
Since I rarely have a tripod with me, when on my travels, some of the image sets are not easy to stitch together, and I’ve noticed all kinds of errors and problems with the output from previous software packages I’ve tried.
Performance issues have also discouraged casual experimenting, since I’m often looking at stitching large 16 bit files, where a minor tweak leads to another 5 minute wait to see what happens.
The Oregon road view below, from several (hand held) shots with a 14mm (rectilinear) lens on a Canon 1Ds (21MP full frame) caused all kinds of problems, both in stitching and in getting a projection I was happy with – I’ll come back to these issues later.
I’m also an architectural photographer, where I most definitely do use a tripod. At the end of last year acquired a GigaPan motorised panoramic head, to enable me to create even higher resolution images (I also have a manual Manfrotto 303sph pano head for taking single row sets of images)
After some initial experiments with the supplied GigaPan software, I realised that a more robust and functionally capable piece of software was required.
There is a lot more such software around now than when I looked a few years ago (see my 2006 review of RealViz Stitcher for example).
I should at this point mention two things that drive my use of such software:
One is that I spent many years, prior to my photographic career, doing research into usability and human factors – one legacy of this, is that I have a very low tolerance of poor interface design and other usability issues. I use software like this as part of my work, so ease of use and flexibility trump cost issues (almost) every time. As such I usually rule out anything that involves command lines or scripting. Others may be fine with it, but I’m looking for software that is both easy to use and does a good job…
Secondly, I’m primarily interested in making great looking photos, not the maths and mechanics of stitching images (although part of me is curious about this too). So, just because an image is made with an impressive amount of detail or a technical feat in itself, doesn’t necessarily make it an interesting photo. To me, panoramic techniques offer new lenses and sensors, not an exercise in projective geometry…
The software can be downloaded directly from the Kolor site.
Upon installation, it looks for other software on your computer that it can work with as a source of images to stitch.
In my case it only found Adobe Bridge (I don’t use Lightroom or Aperture)
The following screen shots I’ve used, all give an idea of the default program layout. It is however quite customisable if you find you prefer a different look.
There are a lot of different icons for different features, which like any new software can take a while to get used to.
Detailed help is always available on-line, or in outline form in the program.
If you are experimenting, then start simple, and assume that if options don’t mean anything to you, then the defaults will probably suffice to start with.
Selecting files and finding matches
The images below show the basic simple workflow, that I’ve found works a treat with many sets of images.
At the top left is a panel where I’ve just dragged and dropped the source images. To the right of it is a panel showing the detected panorama. This is fully automatic – I just pressed the ‘Detect’ button and a short while later a composite image appears.
The third panel below shows a different set of images in edit mode, where I can fine tune how the panorama is presented (geometry, lighting colour and many more).
If I’m happy with this, I can then get the software to render the full size image, based on my choices.
That’s it – with a set of nicely overlapping images, taken at the same exposure, ISO and aperture, I’ve a stitched image within a few minutes (less time than it took to select, adjust and process the 8 RAW camera files for the image).
It would look a bit neater if I’d used a tripod for the original shots, and the colour is a bit odd, because I’m working in the ProPhoto colour space (easily fixed – see later).
Here’s a much reduced version of the output file, with a more realistic version of the sky at over 10,000 feet in Colorado.
To give you an idea of the sorts of resolution you get in images like this, here’s the photographer on the left hand side of the image above.
Picture is a 100% crop from the stitched image.
There are many nuances and adjustments in workflow that you can use to make better source images to start with.
So, I might choose to process my RAW files including lens corrections for distortion and vignetting.
If I know I’m converting an output file to black and white, I’ll certainly be outputing files at 16 bit and in a large colour space.
Suffice to say, the better your source files, the easier it is to make a good quality panorama.
I’ll cover some of the adjustments and fixes that Autopano Giga includes, but my aim is always to use the best source materials.
Whilst on the subject of source files… in 64-bit mode on my Mac Pro (24GB of memory) I’m limited to some half million pixels wide prints (145 feet wide @300 ppi).
There are many options available in the program’s settings, covering how it looks for connections between shots, then how this is optimised, formed into a panoramic image and then rendered as full size output.
It’s worth noting that the software was capable of making ful use of the performance of my 8 core Mac (16 equiv. processors) and makes use of the GPU in my graphics card.
At all stages of the panoramic creation process, there is a lot of information that you can check and further adjustments available.
The example below shows an image stitched from 65 images, where I’m checking the overall image histogram – I could alter things at this stage but I’m trying to minimise adjustments that might be better carried out in Photoshop afterwards.
The info pane is showing me that these images are matching up particularly well – the RMS number should be as low as possible – look for red flags here to warn of potential issues in the stitch.
Some more settings – this time for the way the software optimises connections it’s found between images.
It’s normally fine to leave everything set at default, but once you get further into using the software, some of these adjustments may become important, but don’t worry too much if they remain a mystery – the software mostly just works…
If it all looks a little confusing, there are a number of video tutorials on the Kolor site, and for those like me who much prefer text to read, there is plenty of written help and a discussion forum.
The collections of pictures I’m using in this review are all pretty easy for the software to stitch – if you add in manual editing of control points (the linked common features of images) and the ability to force all images to be included, then the software becomes even more powerful.
There are a number of more advanced functions that I’m not covering in this overview, that are of definite interest to me.
This includes the HDR functionality and the ability to include sets of images with multiple viewpoints. I’ll address some of these features at a later date, once I’ve used the software a bit more.
Plugins and importing
Before creating your panoramic image, you need to get images ‘into’ Autopano Giga.
If you’ve the files already created externally you can just drop files onto the main window (or load a selection from the menu).
With some third party image processing software, there are plugins that allow you to directly export files to Autopano Giga.
In the shot below I’m directly exporting a collection of 65 images from Adobe Bridge – this is fine if I want to build my panorama from a set of JPEG images – but most times I want to use 16-bit TIFF files to maximise image quality and flexibility for subsequent editing in Photoshop.
I note that Lightroom export is a bit more flexible in file export options, but I don’t use LR at all…
Other plugins offer adjustments of images, such as the ‘Neutralhazer’ below, which cuts atmospheric haze
You can see it applied to only one set of images below.
Very much a feature that I’m inclined to say – ‘not now’ to.
But… that’s because I’m used to editing and correcting images in Photoshop, and have lots of tools and filters I’m regularly using. So, if it works for your needs, then by all means try it.
If you looked carefully at the export from Adobe bridge above, you can see that I’m exporting 65 files.
The collection of images was acquired with our GigaPan motorised panoramic head.
I’ve some brief notes covering initial use of the GigaPan
- GigaPan Epic Pro – first look
- Out at night with the GigaPan Epic Pro
- Big prints with the GigaPan Epic Pro
Fortunately, there is a different way of importing these images, through a specialised import ‘wizard’ that supports several specialist panoramic hardware setups.
Here are those 65 images selected via the GigaPan import option.
Fortunately you get a preview (below), which rapidly shows up any duplicates, missing files. or even if you’ve got the number of rows and columns wrong.
A set of night time shots of Leicester Cathedral.
Large areas of black sky like this can cause some stitching difficulties, but rather than reprocess the images in a way that would bring out noise and glow from street lights, I decided that fixing black sky was easier in Photoshop afterwards.
The screen shot below shows that quite a few images failed to link into the panorama.
It was a starry night, so I could have used HDR techniques to get detail in the sky, say three bracketed shots per camera pointing position (More on HDR at Kolor) In general, I try and avoid tone mapping and anything that falls remotely into the lurid/obvious HDR – some seem to like it, but I don’t.
However, this is an architectural shot and with longer exposures, it was already going to take a while to shoot.
Each shot, of the 63 used, was a 6 seconds exposure (f/9 @100ISO Canon 1Ds3 21MP. Since I’m looking to keep noise down, each shot also had a 6 second dark frame shot added (a noise reduction setting option for the camera). If you allow for time moving the camera position, then this whole sequesnce took around 20 minutes.
It’s worth noting that when creating your final images, it’s possible to export files in layered formats that preserve the component parts and stacks of similar shots (same exposure for example), as well as the blended image. I’ll not cover this aspect in this overview, but it allows for considerable fine control of elements of images that may not always blend so well (people getting in the way, being the most common issue).
Next, I’ll look at what you are going to do with your set of images, now that they have been matched together.
First however, it’s worth saving things. This makes it a lot easier to come back to work in progress, or decide on different output options at a later date.
Projections and views
If you’ve never used really wide angle lenses or stitched images into panoramic views, then the whole idea of projections can seem a little difficult to grasp.
Whether you intend to view your images on screen or as prints, you are placing a representation of a three dimensional scene onto a flat surface.
For normal camera shots, there is no great problem – you might notice that when you point the camera up at tall buildings, they look pointy and vertical lines are no longer parallel, but this is hardly something most photographers are concerned with.
As you move to wider and wider lenses, various parts of the picture start looking distorted, particularly people’s faces in the corners of images. Then there are fisheye lenses, which although people think of them as distorted, are actually just another way of representing what the camera sees and puting it onto a flat view. [See my review of Canon’s latest fisheye lens for more examples of this]
Think of the real 3D world around you projected onto a flat plane – the way this is done (it can involve some complex looking maths) alters the look of the final image [See also this video from Kolor]
I’ve no real interest in full 360 degree panoramic images, that cover looking directly up as well as down, so in this review I’m sticking to the sorts of images that I might want to print.
Autopano Giga does support many more types of image, and there are also software packages available from Kolor that specifically address 360 virtual tours [Panotour Pro].
I’ll show some examples of a view of a small park called the Oval on New Walk in Leicester.
BTW If you wondered why a view from this point? …there is a large print, based on this image, in the offices of a business on the other side of the park. Remember – I also do photography to earn a living.
The image has a horizontal field of view of some 210 degrees – far more than any normal lens could cover.
The angle of view is given at the bottom right hand side of the image window.
It’s not that important here that you know exactly which projection works best – just concentrate on the differences as you move your mouse over the following images.
Think of how your image will look when you crop it to the part you want to print (remember to make panoramas bigger than you need – not many people want prints with curved edges).
This second image is also in Mercator projection and shows the cylindrical projection when you mouse over it (note the changed icon in the bar – the dropdown menu is just there because I mis-timed the screen shot!)
This really doesn’t work well for extreme wide angles of view – the centre part of the image does look very good, and is what you’d get with a conventional wide angle lens.
There are lots of tools for altering and fine tuning the view.
For example, with a planar projection, I’m able to set both vertical and horizontal lines in this view of Leicester Cathedral.
The result is a perfectly square – just the sort of view I need for some of my architectural work, and printable at 60″ square at 300ppi.
Choice of projection also makes a lot more difference for a shot like this with a lot of sky in it, shown in a spherical projection (24mm lens on Canon 1Ds3).
I should point out that if you are going to be doing many hand held panoramic shots, you will probably need to become efficient at using the cloning tool to fill in the odd missing bit of an image when you crop to the final size. If you have an issues with this, well, either get over it or remember to shoot more shots than you need for the coverage you want.
There is lots of information available after I’ve stitched the image.
With the fairly static views I’m creating here, taken with good overlap, I’m not having to go in and manually adjust the links and settings for each picture.
The software handles ghosting very well too, this is where something only appears in one image – if you get problems with this, then do check out more of the tutorials and help at the Kolor site.
The panel below the image, shows details of the results of the program’s attempts at matching images and and stacks (groups of similar images)
The software normally works out focal length for each image – this can actually vary, even with the same lens. You can force the software to use the exact same settings for each shot, but the slight variation shown here (fractions of a millimetre) didn’t seem to make any difference to the quality of the image.
Where available from EXIF data, the camera will make use of its camera database to aid its decision making.
This worked fine, although I did get some panoramas incorrectly decide that my EF14 2.8L II lens was a fisheye, and require correction.
The viewpoint and horizon can be altered to get an image the way you want (including setting verticals, as mentioned earlier).
There are automatic adjustments available for image colour and brightness, and individual component images can be altered (more about image corrections at Kolor).
There is an option to use known lens correction parameters if you wish (from the Adobe lens corrections information) although once again I’d generally prefer to do this sort of correction when processing my RAW files.
The blending of images is very efficient and the software is capable of handling quite a bit of variation of source materials.
Once you decide you are happy to produce an output image, then just select ‘Render’ to put the rendering job into a queue.
The shot below, shows four panoramas I’ve assembled, that are ready to go. The program handles multiple tasks at once, so I can easily set images to render, and carry on editing other groups.
Output files can be huge, so it’s worth noting that you can easily specify a smaller size in the rendering options.
Once again there are a lot of options – fortunately the rendering is very quick, so experimenting to see what works is not too much trouble (only a few minutes for multi gigabyte output files)
Do check the default path for your output files, since I temporarily ‘lost’ several, when they were dumped onto the server that I’d loaded the source files from. As mentioned earlier, the output files can contain a lot more than just the final composite image (but can become vast if you are not careful).
The histogram like size slider shows the amount of detail in the output file at different scales. In the example below, I can see that there is a lot more detail in the file than will be preserved if I reduce the output to ‘only’ 1 Metre high.
Depending on how much time you’ve got, there are different ways of resizing – once again this depend on what you are using the output image for.
If for example you want to make a zoomable large panorama for web use, then you may see a difference – for a large print (which I’m going to selectively sharpen for print anyway) this may be less noticeable.
There are a variety of different output file types. The .psd (or .psb) format for Photoshop is my most likely choice, especially for very large files where you may run up against maximum pixel width/height limits.
Don’t forget your disk space either.
Each of the source files (TIFF from RAW files) was 126MB and the ‘Oval’ image when assembled, came in at ~13GB
One minor glitch – you can’t set the output colour profile for the PSD/PSB files, so this warning comes up when opening the file.
Since I know that I processed the RAW files into the ProPhoto colour space, I can simply assign a profile. Remember that these profile warnings are not enabled by default in Photoshop, so do check, if you are using specific profiles.
ProPhoto at 16 bit is useful for holding a wider range of detail and colours and lessening the likelihood of any channel clipping. I often use it when I’m going to be converting an image into black and white (see my review of Nik Silver Efex Pro 2 for more about the nuances of producing digital black and white images).
There is lots of info available after rendering – this is from the cathedral image.
I’ll finish off with a look at two lens types that require a bit more care to get the best from them when stitching.
Wide lenses and Shift lenses
As you could see in the very different projections earlier, stitching wide angle lens images together needs some care and subsequent thought as to what looks best when you crop to a print.
One of the thing I now remember, when taking hand held sets like this, is to take extra shots at either end of what I’m after.
In the pair of images below, extra shots at either side would enable me to make more of that cloud in the top left, when cropping to a rectangular print, whilst a bit more care with working out the real horizon level would have not lost the foreground on the right.
Colour looks a bit wrong due to use of the large ProPhoto colour space.
I know the obvious answer is to do it ‘properly’ with a tripod, but I didn’t have one with me…
At least normal wide lenses stitch pretty easily, but I regularly use another type of specialist type of lens for my work.
The photo hasn’t been ‘fixed’ in Photoshop – this is the image that comes out of the camera.
Note: If you want to know more about what you can do with lenses like this, I’ve written an introductory article about tilt/shift lenses that may be of interest?
The lens in this case is a Canon TS-E17mm, one that I use quite a lot for my architectural work.
What about stitching a horizontal series of shots like this?
Autopano Giga managed some good attempts at stitching, but even with tweaking the accuracy of detection and increasing several other settings, the results were a bit hit and miss.
One other problem with such shift lenses is that you can’t easily apply corrections to fix chromatic aberration and vignetting.
There is however a fairly simple fix – as long as you noted down the amount of shift you used when taking the photo.
The camera was actually level when I took this photo, it’s just the lens shifted upwards.
Think of the image you can see as just part of a bigger one that is missing the bottom part.
Move your mouse over the image to see this ‘lost’ part.[Note original was in error – now fixed]
The green lines show the centre of the visible image, whilst the red lines show the optical centre – a line perpendicular to the camera sensor.
The distance between is the shift.
How do I know how much to extend the canvas downwards?
It’s simply twice the shift, or the number of pixels that represent this much movement for your sensor.
You just need to get used to noting the amount of shift used and know the exact physical size of your camera’s sensor (the 36mm mentioned here applies to my ‘full frame’ Canon 1Ds mk3).
It’s easy to work out and automate for your source images (see more about the calculation and fixing lens aberrations in the tilt/shift article)
You can see the result of stitching several similar shots below.
I’ll have to crop off the black area later, but this canvas extension trick makes it really easy to stitch images taken with shifted lenses.
I’d normally think of using the GigaPan head and a longer lens for such shots, but on this occasion I was using the Manfrotto 303sph head and wanted a better chance of getting some ‘whole’ people left in the shots.
The software does an awful lot more than I’d normally have use for (see the specs below).
In general I view it as providing ‘super lenses’ that don’t (can’t) exist for my camera, and offering amounts of resolution far beyond any commercially available digital camera.
The choices of projection with Autopano Giga provide a useful variety for the sorts of shots I’m likely to take, although I would like to see a few more options with the planar projections, such as some of the modified rectilinear versions I’ve seen used in some images (transverse Mercator or compressed rectilinear). Not a big problem though.
It’s worth remembering at this point that Autopano Giga has a lot of features in it to deal with what I would regard as rather sloppy sets of images – different exposures, apertures, focal lengths.
My general view is that if you want high quality images worthy of making big prints, then you should use good lenses and photographic technique to capture the best photos to start with. The big advantage of Autopano Giga working well with poor quality images is that it works exceedingly well when you have taken care in acquiring your photos – even my hand held sets of wide angle views were no trouble for it (OK, I was using £1600+ lenses ;-)
The blending and stitching worked very effectively with the images I tried – there was rarely any loss of sharpness or detail, and the resulting files were as easy to process, as if they were single shots.
The images below (Loveland pass and Dillon Reservoir in Colorado) were from hand held sets, taken on the same day with my old Canon 1Ds (11MP 35mm full frame) in 2004.
It’s only relatively recently that advances in RAW image processing enabled me to produce really good quality source images for these panoramas, which have quite noisy underexposed areas in them, resulting from my desire to keep detail in the bright sky areas.
Autopano stitched them without difficulty – easily providing source files for making prints over four feet wide.
I found this software particularly easy to use and very rapid in performance.
If you were completely new to panoramic imaging then I’d suggest following some of the examples on the Kolor web site.
Whilst, from my previous engineering and scientific background, I have a knowledge of much of the maths and processing behind what’s going on, I was really pleased that I didn’t need to know any of this stuff to actually use the software. At no time was I presented with a form or dialog box full of meaningless numbers and asked to set some parameter that meant nothing to me. The details are there in the background if I want to delve and experiment, but the whole point is that I didn’t have to know this stuff to start with.
Just because a panoramic picture has gazillions of pixels, doesn’t actual make it a good photo – something I suspect some panoramic photographers are inclined to forget. You still need to think about composition and what your picture is about.
I’d also recommend the user forums on the Kolor site – there is a lot of useful info and knowledgeable people there.
The videos available on Kolor’s site do cover a lot of material, but I’d suggest getting a native English speaker to re-record some of the voice-overs, which I found slightly unclear in parts.
Although RAW processing support is available, there is no adjustment of image processing parameters, so I’d prefer to do any RAW conversion externally, particularly for images such as went into the two shots above.
As someone with previous experience of such software, I was able to get good results without even looking at the manual pages – always a good sign in my usability evaluations.
If you’re searching for panoramic software and want something that largely ‘just works’ then give Autopano Giga (or Autopano Pro) a try.
- May 2012 Making of the large 14m Leicester print Keith’s May 2012 exhibition in Leicester features a single print (below) just over 14 metres long. This article covers all aspects of creating this image, from Keith’s initial experiments with panoramic photography, through to the problems of printing and mounting such a photograph.
- Feb 2013 – APG 3 review . Wider range of projection geometries and noticeably faster.
Full Feature list (from Kolor)
- Detection of about 400 image formats
- Support of RAWs of most manufacturers
- Support of 16-bit images
- Panotools projects import
- Unlimited input resolution
- Fisheye lenses support
- Support of GigaPan and Merlin panoramic heads + all Papywizard-compatible heads
- Automatic detection of images belonging to a panorama in a photo stock
- Ruggedised detection: images can be shot in one row, several rows, skewed, vertical, horizontal, sloping, macros, with telephoto lens, hand-held
- Multi-focal detection: images of different focal lengths can be stitched in the same panorama
- Multi-core support during detection stage
- Bracketed images organized in stacks
- Real-time preview of changes in the panorama editor
- Exclusive algorithm for automatic colour and exposure correction
- Automatic or manual adjustment of colour blending
- Automatic or manual adjustment of colour levels
- Automatic rendering
- Smartblend ghost remover to remove moving objects from the panorama
- Multi-core support during the editing stage
- GPU processing to enhance program speed
- Vertical lines tool to straighten a panorama
- Control point editor for manual stitching and manual corrections
- Automatic or manual cropping of panoramas to an optimum size
- Automatic or manual choice of the best projection mode: planar, cylinder, spherical, Mercator
- Group images per groups depending on their aperture, speed, ISO or focal
- Standard HDR support
- JPG, PSD/PSB, PNG, TIFF, HDR, EXR, Kolor Raw output
- PSD, PSB and TIFF outputs can include layers for direct use in Adobe Photoshop
- Batch rendering
- Unlimited output resolution, creation of gigapixel panoramas
- Choice between 7 interpolators and 3 blending options
- 9 languages included: Chinese, English, French, German, Polish, Russian, Spanish, Italian, Finnish
- Mac, Windows, and Linux compatible
- 32 and 64-bit versions
- Integration in Windows Explorer
- Drag-and-drop interface
There is a version of the software Autopano Pro available at less cost, with a few less features than Autopano Giga, which may well suffice for many users.
|Features||Autopano Pro 2.6||Autopano Giga 2.6|
|Automatic creation of professional quality panoramas (Check the list of common features in image stitching)||Yes||Yes|
|Manual Editor of Control Points||Yes||Yes|
|Creation of gigapixel panoramas and support of motorized heads||Partial (Professional Clauss Rodeon panoheads are not supported in Autopano Pro)||Yes|
|HDR : automatic fusion||Yes||Yes|
|HDR : customisable fusion||No||Yes|
|HDR for special effects (saving of the .hdr file)||Yes||Yes|
|Management of Plug-ins||No||Yes|
|Neutralhazer Anti-Haze Algorithm (plug-in)||No||Yes|
|Support of Adobe Lens Profile to correct lens defects (plug-in)||No||Yes|
|Export of image stacks for processing in third-party software (plug-in)||No||Yes|
|Export plugin for Picasa||Yes||Yes|
|Export plugins for Aperture, Bridge, Lightroom||No||Yes|
|Multiple viewpoints support (for linear or aerial photography)||No||Yes|
|Creation of stitching templates||No||Yes|
|Available in 9 languages||Yes||Yes|
|Compatible Windows, Mac, Linux||Yes||Yes|
Easy to use panoramic stitching software that is both very quick and effective. Comfortably handles large numbers of images such as produced by motorised panoramic heads.
Very broad range of input files supported – mixed focal lengths are no problem.
Works on Mac, PC and Linux
All the latest articles/reviews and photo news items appear on Keith's Photo blog
Keith explains tilt and shift lenses
Keith has written a book that looks at the many ways that tilt/shift lenses can benefit your photography from a technical and creative point of view.
There is also a specific index page on the site with links to all Keith's articles, reviews and videos about using tilt and shift.
We've a whole section of the site devoted to Digital Black and White photography and printing. It covers all of Keith's specialist articles and reviews. Other sections include Colour management and Keith's camera hacks - there are over 1200 articles/reviews here...
Articles below by Keith (Google's picks for matching this page)
We're an Amazon.com affiliate, so receive payment if you buy via Amazon US