-
Posts
504 -
Joined
-
Last visited
-
Days Won
91
cdschroer last won the day on September 13
cdschroer had the most liked content!
Reputation
176 ExcellentAbout cdschroer
Contact Methods
-
Website URL
http://culturalheritageimaging.org
Profile Information
-
Gender
Not Telling
-
Location
San Francisco, CA
-
Interests
All aspects of computational photography. Adoption of new tools and technologies and training programs. Software development of tools for RTI, Algorithmic Rendering, and related technologies.
Recent Profile Visitors
The recent visitors block is disabled and is not being shown to other users.
-
Hi Alice - We don't have such a thing at the moment - but we are figuring out what additional materials we have time and budget to create. I'll note there is an example of using OpenLime with .relight files - on this website of Judean Pillar Figurines: https://laurmcco.github.io/judeanpillarfigurineexhibit/ All the available RTIs can be seen from here: https://laurmcco.github.io/judeanpillarfigurineexhibit/rti.html Our team did the imaging for this project - including the two 3D models using photogrammetry - but we did not do the website, so I'm not a lot of help in terms of how that was done. You can view the page HTML source here: view-source:https://laurmcco.github.io/judeanpillarfigurineexhibit/rti.html
-
Somehow when I read this the first time, I missed that you were using the WSRP viewer which is the only RTI viewer (that I know of) that supports two lights. It was done as proprietary software many years ago, so we (and others working in this space) have no way of knowing how that feature was implemented - and therefore can't support it in any way. If you use the more standard RTIViewer with PTM or RTI files - which is how I looked at these - I couldn't tell any difference. Also, I don't know what your photoshop .jsx script approach was. The only software we've ever supported is RTIBuilder - and now Relight. We have done some batch processing using the PTMfitter in a command line way - so maybe that's what your script was doing?
-
cdschroer started following Blank after loading , Issues with RTI output quality , Relight outputs to generate 3D Wireframes? and 2 others
-
It's really impossible to tell what is going on without seeing a couple of example input images (for example, a low angle and a high angle image) and also a screen shot or two from the RTI to know what is being experienced. Are you creating PTMs or HSH based RTI files? Also, are adjustments being made in the RTIViewer in specular mode - such as turning down the specularity? Frequently when the specular mode is too dark, it's an indication that there isn't enough even illumination across the surface of the subject. Also, there might not be enough stand-off of the lights. How big is the dome? (I mean the radius from the center of the subject to a light?). How big is the subject? You need an absolute minimum of 2 times the diameter of the subject in light stand off. And 3 or more times is better. Also underpowered domes with very small lights can cause issues. If you are turning up the iso on the camera due to under powered lights, that can introduce noise. So can a long exposure. The impact of these things is very dependent on the camera sensor and the noisiness for various scenarios. You can certainly try a different algorithm, but if you have problematic images (which you may not - I can't tell from this description) then a different algorithm won't fix it. It is possible there is an error in the fitting being done in RelightLab - if you can try rebuilding an image set that you built in RTiBuilder with Relight Lab and compare results (using the same algorithm to reduce variation - i.e. HSH or LPTM) then that would be useful info for where the issue might lie. Carla
-
From a single camera point of view (which is what we have with RTI data sets) we can't properly create a 3D surface. We can create a normal map and other maps - from that 1 camera point of view and a series of known lighting positions. While there are approaches to do this, (create 3D from RTI data, or from the produced normal maps) an error accumulates when the normals are turned into a mesh. There are many technical papers about this going back over a decade. Sometimes this is referred to as the "potato chip" effect, because the mesh is warped, like a potato chip, from the accumulated error. If you need a mesh - we recommend using photogrammetry. You can shoot RTI and photogrammetry of the same area, and register the RTI to the photogrammetry data. There are some advantages to that, though there are not currently tools that we know of that can take full advantage of such integrated data sets. There's a recorded talk about combining RTi and photogrammetry from the Symposium we co-hosted at the Met back in 2017. It's Mark Mudges talk on Day 1 in the afternoon https://www.metmuseum.org/about-the-met/conservation-and-scientific-research/projects/rti-symposium/day-1-pm - All the talks are listed for each session, and what time they start within the video. This one starts at 1:11:50 All the recordings from the 2 day event can be found here: https://www.metmuseum.org/about-the-met/conservation-and-scientific-research/projects/rti-symposium Carla
-
Add a set of sphere images into a set of images
cdschroer replied to Aaron Graham's topic in RelightLab
You can use the images with the spheres to make a light position file in RelightLab - and then export it. Then you can bring in a light position file to use with the set of images it applies to. Trying to add sphere images in photoshop will add error to your data in addition to being time consuming. People with domes often use spheres to get the light position file for that dome. Note that this requires that the position and orientation of the camera relative to the dome doesn't change compared to the set of images where you want to apply that light position file. It's OK if the camera moves up and down - like on a copy stand, but don't rotate the dome or the camera - or if you do, then make a new light position file. Under the file menu use save LP and Load LP to accomplish this: I highly recommend naming your LP files in a useful way so that you know when to apply them to what image sets. Carla -
Hi Folks - we generally shoot the scale bars sufficiently next to or resting on the subject - depending on the subject - and then remove them from the scene. So the scale bars are in a total of 12 - 25 images, which we manage and name so that we can easily find them and turn those images on and off as needed during the processing. Here's a link to a time lapse of shooting an outdoor kiosk as part of a training class. You can view the video on the dropbox without downloading it. In this case, there are some scale bars set leaning against the building (it isn't delicate artwork). And another set laid out on the ground next to the building. This was for practice during training - and you do NOT need multiple sets of scale bar to set scale on a subject (though you should have 3 or more scale bars in your scene). After they were shot, they were removed, and the whole kiosk was shot. I think this illustrates a couple of options for setting up scale bars. The scale bars are not casting reflections on the subject in any images that will be used to build final texture maps. The scale bar images are helped in the alignment because of the background image printed on the bars. (i.e. you get a lot of key points and tie points on the bars themselves - which can help your overall project) One note about the scale bars on the ground is that a set of images was taken going over the top of the scale bars in strips, then tied into images that include part of the kiosk by tilting the camera 15-20 degrees at a time (and changing the location of the camera - you never want to just rotate the camera in place like a pano). The distance to the subject for the primary shooting was already determined, and that distance was maintained for shooting the scale bars so no focus changes are needed. For turntable projects we have stands or magic arms, that let us put the scale bars next to the subject, and take those photos by moving the camera and not turning the turntable. Here's an example of a setup: The scale bar holder with the bars is removed as soon as the scale bar shots are taken and before the turntable is used to create circuits.
-
Our scale bars are based on a design from Tom Noble and Neffra Mathews at the US Bureau of Land Management. They had made various scale bar designs before this one, and this one includes a gray scaled background photo. That image is spread across a whole sheet of bars - so you can't make out the full image in a single bar. The image is there to give the software more interest points (key points) to match on, so that the scale bars can be modeled, and may also aide in project alignment. We usually have the scale bars in a separate set of photos next to or in front of the object and then those images are "disabled" once the project is fully optimized and the scale bars have done their job. This also ensures that the images with the scale bars have good look angles with the camera perpendicular (or close to it) to the bars - which reduces marker error. Oblique look angles on the scale bars will increase the marker error - and while you can remove those projections, it's extra work to do so. So, in our general work flow (and yes, there are exceptions) the scale bars wouldn't be included in the photos used for building the final model or texture map, and therefore I don't think they would be seen by Kintsugi. Our scale bar product is pretty low-volume - we sell 40-60 sets a year. And our vendor does not have the ability to print higher resolution targets. We ran tests with their equipment and the materials we need to make them weather resistant - and the bars we sell are our best option - especially to keep them at the current price point. Carla
-
Sorry - didn't get notified about this post. It does not belong in this topic - which is about PTMFitters. There are multiple posts on the site about the "unknown Error Detected" which, most commonly is due to spaces in your path names (file names and/OR directory names). Please teach for "Unknown Error Detected" on the site. Or - better yet - switch to the new RelightLab software as there are many issues and limitations with RTiBuilder. This is from the RTIBuilder download page: Notice! RTIBuilder is old software that does not run on modern OSes. RTIBuilder is being replaced by the open source RelightLab software being developed at the Visual Computing Lab in Pisa. CHI is collaborating on this project, which is currently in active development. Relight website: here Relight Github: here Instructional Video:
-
Sorry - I just saw this. I don't seem to be getting the notifications I used to get from the forums I'm not sure what you are trying to do, but here are a couple of notes. You can convert an existing .rt or .ptm file to the new .relight format. You do not need the original image files to do this. If you want to rebuild from old RTI data, then you would need to start a new Relight project. Relight does not recognize the ole RTIBuilder log files. But the process is a lot faster and more streamlined - so it doesn't take long. You would start a new project, then load the Jpeg-Exports folder. set up the spheres, then detect the highlights. Then export the RTIs.
-
cdschroer started following RelightLab guideline , Blurry RTI Files , How to create a dome calibration file in Relight? and 1 other
-
The size of the spheres, should Ideally be 250 pixels in diameter to get the most accurate light positions. But having smaller spheres than that wouldn't impact the resulting RTI in a visible way - would just mean less accurate light positions. If the area of the subject you are interested in is sharp at aperture 5 - that's what matters. Aperture choice is going to be based on getting sharp focus for the subject in your setup. Sometimes the power of your light can dictate aperture choices as well. I have seen the dome you have, and the LEDs are not very powerful - so going to a wider aperture, and/or taking a longer exposure may be necessary. If you are using a longer exposure, or a high iso - that could be introducing noise - which would impact the result. Zoom a photo up to 400% and see if you are seeing any visible noise. Carla
-
Hi Peter - how did you check for movement? This is the most common cause of blurriness in RTIs - if the images themselves are sharp. Even the movement of a couple of pixels will affect the result - and the pixels on the sensor are only a few microns. Working on a sprung floor for example, can create movement when you walk nearby. A table that is not totally steady, or a tripod or camera stand that allows for even tiny movement of the camera can cause problems. A way I like to check is in RelightLab - after selecting the spheres - then I check the first image and the last image. The circle around the spheres should still line up. This will tell you if you had any movement of the camera relative to the scene. If the subject moved, and the sphere's didn't (perhaps it rocked a small amount on whatever base you are using) then this wouldn't tell you - but you could try checking the first and last images in RelightLab - or in photoshop layers. You need a reference to check against - such as a ruler - because the movement could be very small. I know you are using a prime lens - because sometimes the lens can "creep" when the camera is pointing down. This can happen for focus as well. So, make sure you aren't getting any focus change from the first to the last image. If you use the Digital Lab Notebook - the Inspector tool will check for a change in the "Approximate focus distance" which can be found in the EXIF metadata. You can also check that by hand using EXIFTool, or showing all info in Bridge or Adobe Camera Raw - again, I would compare the first image to the last image in the sequence. Carla
-
There is not yet a manual, as the software is still under development. There is a short video that explains the basic functionality. You can see it here: Not all functionality is implemented - for example Set Scale is not implemented. There is some information about the OpenLime viewer with RTI on this page: https://vcg.isti.cnr.it/relight/ Post your questions here and we will try to answer them. Carla
-
You might consider the new RTI formats designed for Web distribution .relight- and the open source OpenLime viewer - which should run in HTML5 enabled web browsers. More info on the .relight format and the viewer here: http://vcg.isti.cnr.it/relightt You can make RTI's in this format using the open source RelightLab - there are also links on the above page - and elsewhere in the forums. Carla
-
Hello - Sorry - I just saw this. Somehow I didn't get notified when it was posted. You can make a screen shot of the normals view - which will save at the full resolution of your underlying data (like any screen shot in the functionality. A better option for you is likely the new RelightLab software from the Visual Computing Lab (same folks who wrote the RTiViewer many years ago). Exporting a Normal map is a function in that tool. You can also open a previously created PTM or HSHS based RTI and convert it to the new formats designed for the web. I think you can produce a Normal map from an existing RTI as well - though I haven't tried that. Check out the RelightLab forum for more info and links: https://forums.culturalheritageimaging.org/forum/47-relightlab/
-
Alex - Both RelightLab and OpenLime are open source - and available via GitHub under the GNU GPL3 license. So if anyone wants to make such things they can. I don't think that's on the list for the team working on the software. It's not a priority for our team as most of the folks we work with aren't using Wordpress. A higher priority for us is a detailed example so folks know how to set it up. Carla