Thursday, March 01, 2018

A Clever Idea on Paper Falls Short in Tests

Imaging Resource publishes a review of L16 computational camera and the conclusion is quite bad:

"After years of hype and teasers, we finally got our hands on one, and suffice it to say, the image quality and performance leave a lot to desired.

...shooting out in the real world, the L16 is pretty much underwhelming on all fronts."

Fine "detail" crop
Light L16 camera


  1. It is too bad the performance is not better. I posted the following about 5 months ago on a DRP forum:

    Today I sat next to Rajiv Laroia, CTO of Light, during the IS&T and OSA Workshop on Small Eyes, Smart Minds (basically on computational imaging cameras). I gave a talk on the QIS yesterday, and he talked about the L16 this morning. First of all, nice guy, no sense of BS. His background (PhD) is not in imaging but in communications Theory (I think).

    Anyway, all the cameras use the same 13MP sensor (1.1um pixel) - basically a camera-phone sensor, with at least a few custom lenses. Honestly, it is a tour de force in computational imaging and a brave experiment in the consumer photography space. I hope they get traction beyond the tech enthusiasts and very early adopters. The camera was bigger and heavier than I expected (he let me play with his personal L16) and I am not sure it would fit in one of my trouser pockets. His talk included a discussion of the way the camera took an image which included motion of the folded-design mirrors and some other Rube Goldberg events in hardware (like each lens focusing independently) and software. It is remarkable in that sense, that it seems to work. I wouldnt want to drop it. I did compare the display screen image of the floor under the table with the image from my iPhone6S, and for that, the iPhone looked a hair better. But no surprise since I think the display is just from one camera in the L16 and the pixels are smaller than that of the iPhone6S camera.

    Anyway, as I said, an impressive tour-de-force. I am keeping my fingers crossed for Rajiv and the rest of Light. If successful, it would open up a bigger door to computational imaging in the consumer space. A door cracked open by Lytro and the less successful Pelican Imaging. Of course dual cameras in the IPhone 8 and other phones represent baby steps in computational imaging and bode well for less ambitious projects.

  2. IMO the company adopted a wrong biz model. Could have focused on producing a 4-6(max) camera module for cellphones, which would have lowered the science experiment complexity, attracted phone makers as biz partners to share the business risks and get to market faster.

  3. The image shows they couldn't solve the parallax problem, decided to blur the artifact areas.

  4. When I met with Rajiv Laroia, CTO of Light, it was clear that he had drank his own Kool Aid. That's never good for a startup. Belief in the product is essential, but not to the point where the shortcomings and challenges are ignored.

    Like Lyto, Light has made the mistake of trying to sell customers on the technology. Apart from a small pool of technology enthusiasts, you can't build a consumer electronics company on the back of that. You need to sell a product based on what it can do for the customer (what problem it can solve, what new experiences it can enable, etc.).

    The benefits Light is claiming are pretty marginal in an age of smartphones, and for most customers would be better served buying something like a Sony RX100, or Panasonic ZS200 (1-inch sensor camera with a 24-360mm equiv. lens).

    Rajiv Laroia also kept talking about how quickly their Kickstarter campaign had become fully subscribed. Again, not particularly relevant to building a sustainable business selling millions of units a year. The internet puts you in reach of a sufficiently-large pool of people (geeks) to buy tens of thousands of copies of any cool tech product. Millions is a different matter entirely.

    I don't even need to get in to the technical challenges (too many for a startup).

  5. What do they say about MTF at a given object FOV compared to a equivalent DSLR set? It cannot be the same due to the smaller pixel?! -dkf

  6. This is a good example of a product idea that was pursued "because the technology made it possible" - but not because it was a valid [commercial] product idea. It's a neither here nor there thing, which in the presence of steady improvement in smart phone cameras, means it's nowhere.

    1. I respectfully disagree. I try to keep an eye on photography chat rooms and the interest in such a product was fairly high. Of course, that was based on the actual product meeting the expectations created by Light's pre-launch marketing, in turn driven by the need to raise funding for the product development. A very compact, DSLR-quality camera with high resolution and low light sensitivity would be a good product. L16 was not very commpact, not DSLR-quality, not really high resolution, and not so great in low light, so of course totally missed this market opportunity. BTW, there is also the issue of in-camera preview which L16 could not accomplish, except by conventional means.

  7. I was troubled by Light's 16 camera design, a huge engineering undertaking. I don't see a progression of using say 3, 5, 7, ..., 16 cameras for increasing improvements over two cameras baseline. As a consequence, they encountered the practical issues of managing these 16 cameras from day one and had to create custom ASIC. I see this sort of approach as a gamble. If I were VC, I would insist on a half or even third way design with fewer cameras as a proof of concept.

  8. Summary: lots of noisy poor quality images, can not sum up to one high quality image. How much VC money was spent to conclude that? This is what's wrong with Silicon Valley slush funds, fueling human endeavours that are, from the start, going down the wrong path. I've seen this over and over myself, unfortunately.


All comments are moderated to avoid spam and personal attacks.