neuroanatomist said:
ritholtz said:
You both are on the same page regarding AF point selection during tracking. Difference is how system is behaving to select initial AF point. Neuro thinks system goes with nearest AF point with subject. You are saying it uses the same logic as used during tracking. Your text from 1DX2 manual is still not clear if it is applicable for initial AF point acquisition. It could be talking about AF point selection during tracking(dpr tests are showing this kind of behavior). I am not sure. May be you can check with Canon.
^^This^^
As stated above, the text that Rishi is quoting from the manual applies to AF point selection during iTR tracking, not to the initial automatic AF point selection. No need to check with Canon, since Canon has already explained the way initial auto AF point selection for the iTR system works. I've posted these statements from Canon earlier in the thread, but Rishi either doesn't understand them, or more likely is ignoring them becuase they contradict the misinformation which he continues to spread.
Wrong again. Purposefully obfuscating to ensure you never have to admit you're wrong, or that, even worse, you were corrected by that Rishi guy at DPReview...
I'm quoting directly from the manual, which says: "The AF point is automatically
selected based not only on AF information, but also the human face and the subject's color information."
Where does it say 'during tracking' there? Nowhere. It talks about how the point is SELECTED. You claim that has nothing to do with how it's INITIALLY SELECTED.
Where's your evidence? Rudy Winston's quotes? Which say NOTHING about how the point is initially selected? In fact, all Rudy describes is how subject tracking works. On, like, all cameras these days: some sort of image sensor is used to analyze your initial subject in order to automatically shift the AF point to stay on it. Nothing groundbreaking here - this is simply how subject tracking works.
I have no idea why you keep quoting Rudy's description of subject tracking. Duh. That's how all subject tracking works.
neuroanatomist said:
Rishi, still waiting for you to:
1) acknowledge that you've read and understood the way in which Canon describes the initial automatic AF point selection in AI Servo, and
2) provide actual evidence that Canon is wrong in the way they describe the function of iTR
Here you go:
Try and focus on someone's face who's holding out a hand in front of him that isn't obscuring his face, but still falls within the AF module.
According to your theory, the camera would always first focus on the hand (because it's closest), and only
then the face. Not only is that stupid, it's not what happens. Furthermore if the hand were far enough in front of the face, the face might fall so far out of focus that the RGB metering sensor wouldn't be able to 'see' the face anymore after the AF system focuses on the nearest subject. So that'd be one
stupid way to program the system.
Thankfully, Canon didn't program it as stupidly as you, and you claim Rudy, suggest (I actually don't think Rudy suggests this at all - I give him more credit than you).
Which you can verify by
just doing the experiment.
So, the fact that the camera
doesn't first jump to the hand and then the face proves
you're wrong, and I'm right, to put it bluntly.
But worse, as I've said
n times at this point, and you've
conveniently ignored, even if you were right, you'd still be incorrect in your assessment of our piece - which concluded that Auto Area was simply more useful on the D5 than the 1D X II. That's true regardless of intended design. Whether or not Canon was DESIGNED to only focus on what's nearest (which it's not, not the least because that'd be a stupid design when you have the information off the metering sensor), if that's what it did, while the other was more intelligent about what it chose, then that'd still be an interesting result -
period.
Let me re-state this for the tenth time:
regardless of what Canon's actual programmed behavior is, how does any of this discussion detract from the fact that our observation was that the Canon's mode was less useful in that shooting scenario?? It doesn't. You're just trying to hide behind 'well DPR didn't even know this was the original intent, so they're stupid, so you shouldn't listen to them'.
That's really mature. And ultimately wrong anyway, because again we're the ones who actually know how it works, proven by the simple experiment I suggested, and backed up by Canon's own wording in the manual.
In fact, ironically, it is you who are suggesting that Canon has programmed their Auto mode stupidly, and us giving it more credit. I'd ask Canon Japan what the actual design were, if only your suggested behavior weren't so stupid (when you have a metering sensor) that my asking if that's how they programmed it might risk me suggesting their engineers are idiots.
Also, they might then just refer me to the manual, that, again for the 10th time, states 'the point is selected based off of face and subject color information'. NOWHERE does it state in that iTR discussion 'but the initial point is the one that registers the nearest point, AFTER which if a face is detected the camera suddenly switches erratically from the originally chosen nearest subject to a face'. Because, you know, that'd be a really stupid way to program your algorithm, and I actually give Canon more credit than you do, it'd seem.
Even worse, your claim that we now love shooting in Auto Area modes and that's what we care about is ludicrous - no one at DPR shoots this way, we just recognize that it can be useful, especially as a backup method you can activate instantaneously to try and salvage the shot if all else fails.
Our testing it is suddenly construed as us thinking it's all-important, and is now weighted into our scoring system heavily, as you and fellow commenters now suggest?
What on earth??
neuroanatomist said:
Note that the following – all of which you've already tried, in some cases more than once – do
not constitute actual evidence:
- restating your opinion
- shouting
- sarcasm
- name-calling
- side-stepping the issue by again talking about how iTR behaves during tracking
- ignoring the information provided by Canon
- use excuses to justify not responding
As I previously stated, I'd welcome additional/new factual information that contradicts Canon's own statements...explicit statements which you ignored the first time, and which I re-posted above. But you seem unable to provide any such factual information. As the saying goes...put up or shut up.
Well, for one, I gave you a perfectly executable experiment that proves you're wrong and I'm right. Do it. I can't wait to hear what sort of justification you have next to prove we're wrong and you were right. It's already amazing you can take the very words written in the manual and twist them to suit your theory because you pull Rudy's words that essentially say the same thing, yet don't at all distinguish what's done 'initially' vs. not to suit your own theory. I believe there's a term for that...
Finally, how do you get on your high horse and accuse me of shouting, name-calling, sarcasm, etc.?
You take a look in the mirror lately? At least I don't even bother accusing you of this stuff, because I expect it from you. But then again at least I have the decency to have this open conversation with my name tied to everything, as opposed to sitting behind a fake username insulting whomever I desire, calling them trolls, biased, incompetent, and putting words in their mouths they never said in hundreds of posts in multiple threads...
-Rishi
