ProfHankD

ProfHankD

Lives in United States Lexington, United States
Works as a Professor
Has a website at http://aggregate.org/hankd/
Joined on Mar 27, 2008
About me:

Plan: to change the way people think about and use cameras by taking advantage of cameras as computing systems; engineering and creatively using camera systems to provide new abilities and improved quality.

Comments

Total: 1699, showing: 1 – 20
« First‹ Previous12345Next ›Last »
In reply to:

ProfHankD: Pentax already tried mirrorless on a DSLR (also SLR) mount -- not a big winner.

JNR: I think they tried, got an answer of "NO," and then moved on. I think both Nikon and Canon have made such a marketing point out of their OVFs that they are also likely to get an initial "no" to any mirrorless using the DSLR/SLR mount.

Link | Posted on May 22, 2018 at 14:06 UTC

Pentax already tried mirrorless on a DSLR (also SLR) mount -- not a big winner.

Link | Posted on May 22, 2018 at 13:31 UTC as 340th comment | 3 replies
On article DPReview TV: Waterproof camera shootout 2018 (173 comments in total)
In reply to:

ProfHankD: You know, Olympus has led this market for a while, but I still think the TG-5 misses a bit. People don't seem to understand that in tough shooting situations, availability of a very wide view angle and flexibility about how you view the display are more important than normal. Instead, we see companies building two lines: hardened point-and-shoots and ultrawide/fisheye action cameras.

In 2015, the Olympus Stylus Tough TG-860 came out with a 21-105mm equivalent lens and a tilting LCD that allows angling the screen so you can see the live view without having your face directly behind the camera (it even allows selfies) -- both of those features are qualitatively more useful than what these other cameras offer. I just wish Olympus had continued TG-860-like models, especially improving image quality....

Well, most higher-end cameras now have at least tilt... so it can't be that hard an engineering problem. Honestly, I doubt it was a ribbon cable that failed; more likely, some humidity got past a seal -- LCDs are super-sensitive to humidity/water. There is admitedly a higher probability of minor leaks in the tilting LCD because it's a thin part that can be slightly deformed by twisting. Anyway, so far, no troubles with my TG-860....

Link | Posted on May 21, 2018 at 15:56 UTC
On article DPReview TV: Waterproof camera shootout 2018 (173 comments in total)
In reply to:

ProfHankD: You know, Olympus has led this market for a while, but I still think the TG-5 misses a bit. People don't seem to understand that in tough shooting situations, availability of a very wide view angle and flexibility about how you view the display are more important than normal. Instead, we see companies building two lines: hardened point-and-shoots and ultrawide/fisheye action cameras.

In 2015, the Olympus Stylus Tough TG-860 came out with a 21-105mm equivalent lens and a tilting LCD that allows angling the screen so you can see the live view without having your face directly behind the camera (it even allows selfies) -- both of those features are qualitatively more useful than what these other cameras offer. I just wish Olympus had continued TG-860-like models, especially improving image quality....

The TG-870 is discontinued with no follow-on; they point at either the TG-5 compact or the TG-Tracker action video camera as the next, which I'm saying is a very unfortunate thing.

Link | Posted on May 21, 2018 at 11:13 UTC
On article DPReview TV: Waterproof camera shootout 2018 (173 comments in total)
In reply to:

ProfHankD: You know, Olympus has led this market for a while, but I still think the TG-5 misses a bit. People don't seem to understand that in tough shooting situations, availability of a very wide view angle and flexibility about how you view the display are more important than normal. Instead, we see companies building two lines: hardened point-and-shoots and ultrawide/fisheye action cameras.

In 2015, the Olympus Stylus Tough TG-860 came out with a 21-105mm equivalent lens and a tilting LCD that allows angling the screen so you can see the live view without having your face directly behind the camera (it even allows selfies) -- both of those features are qualitatively more useful than what these other cameras offer. I just wish Olympus had continued TG-860-like models, especially improving image quality....

mais51: Could be, but isn't on mine. For example, it's great to be able to do things with the camera underwater and my face not.

Link | Posted on May 21, 2018 at 03:30 UTC
On article DPReview TV: Waterproof camera shootout 2018 (173 comments in total)

You know, Olympus has led this market for a while, but I still think the TG-5 misses a bit. People don't seem to understand that in tough shooting situations, availability of a very wide view angle and flexibility about how you view the display are more important than normal. Instead, we see companies building two lines: hardened point-and-shoots and ultrawide/fisheye action cameras.

In 2015, the Olympus Stylus Tough TG-860 came out with a 21-105mm equivalent lens and a tilting LCD that allows angling the screen so you can see the live view without having your face directly behind the camera (it even allows selfies) -- both of those features are qualitatively more useful than what these other cameras offer. I just wish Olympus had continued TG-860-like models, especially improving image quality....

Link | Posted on May 20, 2018 at 22:07 UTC as 40th comment | 8 replies
In reply to:

ProfHankD: Hmm. It's a bit disturbing that they didn't reference our Electronic Imaging 2016 paper, "Use of Flawed and Ideal Image Pairs to Drive Filter Creation by Genetic Programming" -- https://doi.org/10.2352/ISSN.2470-1173.2016.18.DPMI-016 . That paper, which won best paper in its conference, explained in detail how to use machine learning based on pairs of actual captures (of high and low ISO images of the same scenes to serve as flawed and oracle image) to derive image processing filters. Our work used Genetic Programming (GP), rather than restricting the filter design to a neural network.

In any case, the results in this paper look good. However, the paper isn't clear about how it did with images not from the training set. Since the idea is no longer novel, I would have expected more real performance evaluation in the paper....

Yes, I'm talking about the citations in their paper. I find this one particularly annoying because of the press coverage this paper has gotten, but missed citations happen all the time in academia.

Link | Posted on May 15, 2018 at 10:53 UTC
In reply to:

ProfHankD: Hmm. It's a bit disturbing that they didn't reference our Electronic Imaging 2016 paper, "Use of Flawed and Ideal Image Pairs to Drive Filter Creation by Genetic Programming" -- https://doi.org/10.2352/ISSN.2470-1173.2016.18.DPMI-016 . That paper, which won best paper in its conference, explained in detail how to use machine learning based on pairs of actual captures (of high and low ISO images of the same scenes to serve as flawed and oracle image) to derive image processing filters. Our work used Genetic Programming (GP), rather than restricting the filter design to a neural network.

In any case, the results in this paper look good. However, the paper isn't clear about how it did with images not from the training set. Since the idea is no longer novel, I would have expected more real performance evaluation in the paper....

J A C S: I would tend to agree that neural networks are currently overused in image processing and this paper does cite a lot of prior work: 43 papers. I still find it disturbing that our much-more-relevant-than-most work wasn't cited, even if the omission was the result of nothing more than a somewhat myopic literature search.

Link | Posted on May 15, 2018 at 04:39 UTC
In reply to:

ProfHankD: Hmm. It's a bit disturbing that they didn't reference our Electronic Imaging 2016 paper, "Use of Flawed and Ideal Image Pairs to Drive Filter Creation by Genetic Programming" -- https://doi.org/10.2352/ISSN.2470-1173.2016.18.DPMI-016 . That paper, which won best paper in its conference, explained in detail how to use machine learning based on pairs of actual captures (of high and low ISO images of the same scenes to serve as flawed and oracle image) to derive image processing filters. Our work used Genetic Programming (GP), rather than restricting the filter design to a neural network.

In any case, the results in this paper look good. However, the paper isn't clear about how it did with images not from the training set. Since the idea is no longer novel, I would have expected more real performance evaluation in the paper....

It's GP (Genetic Programming) -- creating a filter program by evolving an abstract syntax tree defining the structure of the filter program. Neural networks are a tiny subset of the functional forms it could generate. The catch is that the evolution/training is much more complex for GP, and we didn't give it very much time on our now-unimpressive supercomputer.

Link | Posted on May 15, 2018 at 00:15 UTC

Hmm. It's a bit disturbing that they didn't reference our Electronic Imaging 2016 paper, "Use of Flawed and Ideal Image Pairs to Drive Filter Creation by Genetic Programming" -- https://doi.org/10.2352/ISSN.2470-1173.2016.18.DPMI-016 . That paper, which won best paper in its conference, explained in detail how to use machine learning based on pairs of actual captures (of high and low ISO images of the same scenes to serve as flawed and oracle image) to derive image processing filters. Our work used Genetic Programming (GP), rather than restricting the filter design to a neural network.

In any case, the results in this paper look good. However, the paper isn't clear about how it did with images not from the training set. Since the idea is no longer novel, I would have expected more real performance evaluation in the paper....

Link | Posted on May 14, 2018 at 22:32 UTC as 34th comment | 9 replies

Good article, nice images, but "Earlier this year, with the support of Nikon Europe, I tested it out" sure sounds to me like this lens was a loaner and maybe there was some other support? That sort-of goes against the spirit of "Roie Galitz was not compensated for this article in any way either by Nikon or DPReview, and Nikon had no involvement in its publication." Not a big problem, but perhaps it would be better to more precisely disclose what the "support of Nikon Europe" really meant....

BTW, I think it's good that Nikon is still creating very specialized lenses like this. It might not be a lens "normal people" will buy, but I can easily imagine every newspaper, magazine, etc. equipment pool buying one or two, and it's all the more reason to stick with Nikon.

Link | Posted on May 11, 2018 at 13:01 UTC as 20th comment | 1 reply
In reply to:

Playright: 30 Shots per charge! I think you meant 300!

Yeah; they say 300 in the specs.

BTW, I still think that's a problem. When a camera is wet, changing the battery can get moisture on the compartment seal, which will cause a leak.

The OVF really isn't such a big deal for a rugged camera. I have an Olympus Stylus Tough TG-860 and the tilting rear LCD (which can even tilt all the way up for selfies) is much more useful. Having the LCD tilted, for example to waist-level viewing, really helps hold the camera in places I can't get my face.

Link | Posted on May 11, 2018 at 10:10 UTC
In reply to:

Retzius: I think the genius that decided to polish the inside of the lens hood should get a promotion

Even the Waterhouse stops are gold. Designed for Donald Trump? A Russian lens that gives golden showers of flare. Ew.

Link | Posted on May 10, 2018 at 00:40 UTC

Much as I'm not crazy about Fujis, I have to give them credit for adding some nice features in firmware updates. Sony in particular is really missing this boat. Things like extreme focus bracketing could easily have been implemented via camera apps in Sonys... but the latest Sonys don't support camera apps and don't have these features built in.

Link | Posted on May 9, 2018 at 00:41 UTC as 28th comment | 2 replies
On article Sony 'striping': here's the fix (791 comments in total)
In reply to:

AZBlue: For all the posting Sony fanboys have done in the Nikon forum, this is truly hilarious. Payback is a beyotch! So SO glad I have a Nikon D850. It actually works and takes excellent images that I don't need to fix in post.

Oh, and did you see the "fix" for this striping issue isn't even from Sony - it's from some third party, which is basically a patch job. And the resulting image comes out slightly softer than the original. So much for your Sony's resolution... I'm laughing so hard right now...

Rishi: DFD requires a lot of computational sophistication to work with a single frame sample -- which I don't think the current versions have. However, it's a good bet that will happen eventually -- which is why I say long term. I'm also mystified by Canon's DPAF clipping and AA filtering, but not too smiley-faced about it. :-(

As for the "output version" of light fields, you can definitely go see it working. The company is http://misappliedsciences.com/ and CTO Paul Dietz is my brother. It's very newly public that they are doing this; Geekwire did a little piece on it: https://www.geekwire.com/2018/breakthrough-parallel-reality-display-technology-promises-personalize-world-without-goggles/

Link | Posted on May 4, 2018 at 22:52 UTC
On article Sony 'striping': here's the fix (791 comments in total)
In reply to:

AZBlue: For all the posting Sony fanboys have done in the Nikon forum, this is truly hilarious. Payback is a beyotch! So SO glad I have a Nikon D850. It actually works and takes excellent images that I don't need to fix in post.

Oh, and did you see the "fix" for this striping issue isn't even from Sony - it's from some third party, which is basically a patch job. And the resulting image comes out slightly softer than the original. So much for your Sony's resolution... I'm laughing so hard right now...

Rishi: Canon's DPAF has very serious crosstalk problems, and I have to wonder how much of that is due to an AA filter rather than VLSI fab issues. That said, DPAF is just a very crude version of a plenoptic sensor -- distinguishing just two (not very sharply defined) ray angles per pixel. I'm still not sure if I believe plenoptics are the future. For AF, I think depth from defocus using PSF recognition might be the longer-term winner, and even if you believe in what Lytro was pushing, there are other ways to get essentially the same data plenoptics get you (e.g., coded apertures).

Incidentally, I do believe in the "output version" of plenoptics: lightfield display technology. However, I'm biased. My brother is CTO of Misapplied Sciences, a Seattle-based company making "Parallel Reality" displays that can show thousands of people different things at the same time. :-)

Link | Posted on May 4, 2018 at 02:34 UTC

Super-Wide is less wide than Ultra-Wide?

Well, I guess that's good news for Ultraman and bad news for Superman. ;-)

Also strange they quote field of view for wides and X factors for tele and macro....

Link | Posted on May 2, 2018 at 17:59 UTC as 6th comment | 2 replies

So, we have an NIR/ViS/NIR+VIS camera with 10um pixels doing 720P (1280x720 video) and a fast (normal) lens. Interesting. I wonder if they did something cute like R,G,B,NIR for the CFA? (Probably not.) The sensor does have 2-frame HDR support at 30FPS!

They also have their sensor available as a dev kit, so the specs are here: https://www.sionyx.com/uploads/1/1/8/7/118701572/sionyx_xqe-1310_data_sheet_v08.pdf

I have to admit that the sensor sounds pretty appealing....

Link | Posted on May 1, 2018 at 16:49 UTC as 12th comment | 2 replies
On article Sony 'striping': here's the fix (791 comments in total)
In reply to:

Full Stop: Could someone make a picture where the object has a real thin vertical scratch?
Just to check if we are running into a new "scratch-eater" problem with this fix.
:-)

BlueBomberTurbo: "As stated by the creator," -- just to be clear, that's me. :-)

Link | Posted on Apr 28, 2018 at 18:32 UTC

Ok. This seems a better product idea than the "Digital Rolleiflex" from 2011 that was too expensive and had a tiny 3MP sensor. I would be happier if they didn't call it a "Kamera." ;-)

Link | Posted on Apr 28, 2018 at 01:52 UTC as 15th comment
Total: 1699, showing: 1 – 20
« First‹ Previous12345Next ›Last »