Thursday, November 17, 2016

Lytro Wrap-Up - Summaries and Conclusions

Now, I'll wrap up my thoughts on the Lytro - from my expectations to final workflows and deciding which format to publish.

Hopes and Dreams

When I first bought the Lytro Illum, I had some ideas about how to use it.
  • Post-process correction of stills. That is, fix slightly out of focus photos.
  • Produce something novel - either a "living picture" where the viewer may choose what they focus on or an animation of a rack focus that may have a Matrix-like feel to it. I thought this would be especially interesting with a long shutter motion blurred photo with a flash at the end.
Overall, I thought the "focus later" was going to help me save time by allowing me to be sloppier when taking the photo, and by letting me fix it later.

In Practice

In my last few entries, I've gone into a lot of detail describing the frustrations I had.
The camera/hardware itself was mostly comfortable to use but I did have a few complaints:
  • It was difficult to tell if the camera was turned off or just in sleep.
  • Autofocus was too slow for moving subjects. (dancers)
  • I couldn't always tell what distance it chose to focus on, or what the optimal distance was, given the two peaks.
  • The depth offset and depth bracket controls were confusing.
My complaints were more with the Desktop software to process the image.
  • The processing was slow and resource (hard disk, memory) intensive. For practical purposes, it required a newer computer than what I had available.
  • The processing relies on a depth map which is full of errors. This also meant that I was not going to get my matrix-like shot with the flash at the end.
  • Desktop was cumbersome for organizing photos and was slow to advance to the next photo.
  • Desktop wasn't good about giving me meaningful numbers/units for the depth.
  • To me, it didn't look like the final image reflected the bokeh or focus falloff of a regular lens.
Overall, my workflows generally wound up trying to avoid Desktop and use other tools (Lightroom, iMovie) to process the image/animation. Lytro uses a proprietary image format. I wish that they had provided plugins for Lightroom and iMovie to deal with it, but they didn't. So I need to use their Desktop software. I tried to restrict my editing in Desktop to depth and f/Stop decisions.

And finally, the end result is just a 4 Megapixel image - only half of what my phone can do - and the picture takes about 50MB. Overall, I didn't feel like I could refocus over a very wide range of depths and it wasn't clear exactly what was in focus. Also, I didn't think there was a high enough sample density so it couldn't distinguish small changes in depth.

Target Audience was Unclear.

I couldn't figure out who the target audience for the Illum was. The Lytro website features testimonials from professionals including a wedding photographer so I think they are setting the expectations pretty high.

My question is: what was the last camera they expected their customer to have used?
I have different critiques for these - my feeling for what different people may like or dislike:
  • camera phone - probably a casual photographer
    - $1500 retail is a problem (though after they discontinued the camera, it went down to $400)
    - Carrying around an extra device(the camera) is a problem
    - Probably doesn't have a desktop/laptop with the processing power required.
    - Might be okay with limited control in software (like the animation transitions), but would expect the ease of use and polish of something like iPhoto.
    - 4 Megapixel image may be okay
  • crop sensor SLR / advanced hobbyist
    - probably okay with price
    - used to carrying around the camera, but probably prefers not to carrying around two.
    - Might have the computing power needed for Desktop
    - Likely frustrated with limited control in software. Probably expects feature set of Lightroom.
    - 4 Megapixel image a little frustrating
  • full frame SLR / pro
    - probably okay with price
    - may be used to carrying multiple cameras
    - Should have computing power needed for Desktop (though I don't)
    - Many controls lack precision desired (typing in a number)
    - Expects features, ease of use,  and polish of Lightroom
    - 4 Megapixel image very frustrating. Not enough margin to crop. More than that, having a tiny image sensor meant noisy images indoors.
  • tech geek
    - might be okay with price and second camera
    - Likely has computing power needed for Desktop
    - Probably frustrated by controls that lack precision of being able to type a number
    - Okay with prototype software if there's an open API to access the rays.
    - Frustrated by lack of API for reading the raw file(accessing the rays). (Though I have seen(googled) some reverse engineering.)
    - 4 Megapixel image might be disappointing, but okay
That is, there are different aspects of the camera and software experience that would frustrate different types of users, and no one was fully addressed.

Being an Artiste / Composition

The gimmick of the Lytro is having multiple focus points. This can be highlighted in a couple ways:
  • As a "living picture" where the viewer can click on the image and choose what is in focus.
  • As an animation where I design/compose the transition, and try to use cropping and focus to direct the viewer's eye through the photo. (Discussed in Part 4)
The gimmick works best if there is more than one interesting thing to look at in the scene. Maybe it is one couple dancing in the foreground and another couple or the band in the background. If there are multiple points that do a good job of showcasing a couple depths, I'll probably put together an animation. If the different focus points are really clear, and the depth map isn't too bad, then it may be a candidate for a living picture.

If there is only one point of interest (maybe a couple dancing and only a wall behind them), it may be better suited for a static still. This was the case with most of my images.

It is often challenging enough to get one couple in a good pose, but for Lytro living pictures, I need to time/choose it so multiple couples are in a good pose. Also, if the depth map doesn't have much range, there won't be anything for the viewer to play with.

I've been a little reluctant to pursue the living picture. The only place to present them as a "living picture" where the viewer can choose the focus point is on Lytro's website. I don't really like the user agreement there, but it is the only site that can really host this type of image. Also, it also means surrendering some artistic control of the shot to the viewer(that's you).

This all being said, I did upload a few living pictures for a few of the photos I used in the animation:
Here is a shot that I thought turned out okay. I think it helped that it was an exterior shot and the daylight let me shoot with a lower ISO. I think this made it a cleaner image and helped Desktop compute a better depth map.

If I choose an animation, I get some control back, and I also get tools like cropping to help guide/direct the composition. In an animation, I can animate the crop to simulate a pan or zoom.

At Inspiration Weekend, most of my photos had a dancer from head to toe around the center of the frame. Imagine the frame divided in thirds like a tic-tac-toe board, and the dancers in the middle - their heads a little above the line, and feet a little below the line. There's a lot of variation, but I think that's what I was shooting most of the time.

That usually meant that I was shooting a medium to wide angle, and I may be about 25 feet away from them. For these kinds of shots, the Illum wasn't able to get a good depth map. My speculation is that with that composition, each pixel covered too wide an area in the world. Also, the separation that you might be able to get for different views of an object was too small.

This next shot is from the "Average" gallery. There is very little depth information, so there are not a lot of points to focus on that show a transition. You'll want to click on faces, but I think sometime it may have been shoes that actually landed on a different plane.

From the tutorial videos, it looked like they were getting the best results when the main subject was less than 10 feet away. Also, they seemed to encourage zooming in a lot (longer focal length) so that the close subject is occupying about a third of the frame. But those were not the type of shots that I was trying to get.

A Tool to Fix Stills

Sometimes, I might not use the gimmick at all in the presentation. If a shot didn't meet my requirements for a animation or "living picture" (multiple points of interest at different depths),  I may still want to publish it as a still. I was also hoping that the "refocus later" would be a tool to let me recover some still photos that were slightly out of focus.

When I shoot dance photos, the subjects are often moving and changing depths from the camera. I like the look of a narrow depth of field, which typically means a low f-stop (like f/3.5 or so). Often, when the subjects are a medium distance away and moving, the camera may choose to focus on the back wall or something and the subjects are noticeably soft. Sometimes,  I may set the focus on one dancer's back that's closest to me and the farther dancer's face will be a little out of focus.

In practice, I was disappointed in this aspect of the Lytro Illum. Perhaps, more accurately, I was disappointed in the Desktop editing software. I've discussed my issues with involving depth maps in the process. From the user manuals, it seemed that while the Lytro could refocus over a range, that there were 2 distinct sweet spots. Even under ideal conditions, it would not be effective in fixing something that was just a little bit off.

Future Work

Admittedly, I think I may have run the camera through an extremely difficult gauntlet.
  • Moving subjects (dancers) - autofocus couldn't track, need fast shutter
  • Indoor lighting (dark) - ISO 3200 wasn't up to snuff, though f/2 was good.
  • subjects a medium distance away so each pixel sweeps a bigger area in world space
  • medium to wide shots - there probably was not a wide disparity between different views, which probably made it difficult to compute the depth map.
With that in mind, maybe I could try another round with more shots of head down to shoulders or waist with brighter lighting, such as outdoors. I speculate the camera would work best if the main subject is within 10 ft of the camera, so maybe I can test that hypothesis.

Final Thoughts

In the end, I thought that this may be an okay camera to shoot a few novelty shots as a second camera, but it was not up to the task of being the primary camera at a big event.

If you do want to play with the Lytro Illum, they're currently (Nov 2016) selling for about $300 on Amazon. Just make sure you have the computing resources to deal with it. The files themselves are large - approx 50MB each (quick multiplication: 200 images needs 1 GB). I found the library organization tools in Desktop confusing for someone like me who's used to Lightroom and straight file systems.

If you're using a Mac, you need Yosemite(OSX 10.10) or better to run Desktop 4. But OSX is already up to 10.12(Sierra), so it will harass you repeatedly if you don't upgrade. Newer OS versions tend to overwhelm the resources of older hardware. Also, they recommend 4GB of RAM, and I would take that seriously. I only had 2GB and everything felt slow.

I'm not as familiar with the performance on a Windows computer, but they have Desktop 5 available. Though I think there are specific graphics cards it works with, so make sure you meet the system requirements.

As I was shooting, I had a lot of uncertainty whether I got an acceptable shot or not, and too often the answer was "no." (see above notes on Future Work)

It was frustrating to consume a lot of resources (disk space, memory) and a lot of my time to in the end, produce a mediocre 4 Megapixel image. I think that, in combination with not being able to identify a target consumer, led to Lytro pivoting away from the consumer camera market. Well, that and VR is super hot right now.

No comments:

Post a Comment