Microsoft's Photosynth Will Fail

EddieC 0 Tallied Votes 341 Views Share

In the late 1990s an innovative company called Enroute Imaging came out with QuickStitch, a program that could analyze a series of digital photos and "stitch" them together into a single image. It cleverly figured out where the image data repeated and combined the photos seamlessly. It was really cool software, but I couldn't quite see a wide need. The company fizzled. A number of companies have popped up since then offering similar solutions. Can you name one? I didn't think so.

But now Microsoft has released , a Web service that does largely the same thing, but using its own back-end systems for storage and processing horsepower. The idea is that people would feed in between 20 and 300 photos of a particular subject and Web site would spit out a "synth ," a navigable virtual representation of the images that you can pan and zoom. With the technology, Microsoft claims to turn 2D photos into 3D images, but that's a stretch. I'd describe it more like QuicktimeVR or like taping a bunch of photos to a clear beach ball and looking at them from the inside.

On the client side, a small viewer apps is required, as is an account on , Microsoft's collaborative web site for "scientists, engineers and entrepreneurs," and now photographers, I guess.

It's interesting to note that the Photosynth client does not require Silverlight. The tool relies instead on Flash. That fact could be its saving grace since Flash is already present on the vast majority of systems and isn't limited by Silverlight's relatively large horsepower requirements.

And on the subject of horsepower, the site slowed to a crawl the first day, as it saw 286,000 image uploads, according to an . So much for good first impressions.

Dani AI

Generated

A short, practical addendum to the discussion: ’s instinct that Photosynth wasn’t magic is fair — it didn’t conjure full CAD-style geometry from a handful of shots. Technically Photosynth built on structure‑from‑motion ideas: detect matching keypoints across many overlapping photos, estimate camera positions, and generate a sparse 3D point cloud that lets the viewer jump between original images in a way that feels three‑dimensional. That lineage goes back to academic work like PhotoTourism and the UW / Microsoft research collaborations that powered the early demos. (phototour.cs.washington.edu)

On the product side, ’s point about scale and user expectations is the other half of the story: great research demos often struggle when turned into a public service because of compute cost, UX complexity and the requirement for dozens (or more) of well‑overlapped photos. Microsoft eventually retired the public Photosynth service in 2017, though the ideas and some code resurfaced in other Microsoft photo efforts afterward. That pattern — research → demo → selective product integration — is common for ambitious vision projects. (wiki.archiveteam.org)

If the thread’s interest is practical (wanting the capability rather than debating the company), current tools are far more accessible: open‑source SfM pipelines like COLMAP and AliceVision/Meshroom, and commercial packages such as Agisoft Metashape or RealityCapture, will build dense reconstructions and textured meshes you can export to WebGL viewers. Expect a learning curve and nontrivial processing time (many tools benefit from a CUDA GPU), and follow basic capture rules: lots of overlap, consistent exposure/white balance, and multiple viewpoints. For research background and starting points, look at the PhotoTourism papers and the modern SfM tool pages linked above. (github.com)

acejames1 0 Junior Poster

286,000 that is alot of images most of them came from files that people already had on there computer which is very interesting. so now they know what kinda bandwidth they need but i bet they had a good starting point for the site. but the amount of users probably overwhelmed them. and made by microsoft wow they should be extremely grateful because every product they have come out with they seem to make a screwup out of and noone goes out and buys it. office is a great example of how bad microsoft can screw up and i dont doubt they will eventually get around to screwing up this program

Be a part of the DaniWeb community

We're a friendly, industry-focused community of developers, IT pros, digital marketers, and technology enthusiasts meeting, networking, learning, and sharing knowledge.