Douglas Liu wrote:
I am not sure, some of the footage looked like they were from a TV screen.
While I'm not sure how some of those were recorded, they didn't look like I expect recording a TV to look. In the older sample where people really were recording a TV in a shop, it was very obvious.
Douglas Liu wrote:
I am not sure, some of the footage looked like they were from a TV screen.
It looks like it was recorded off the LCD with a cell phone and according to the written review it was all taken in the wild (ie zoo). It wasn't recorded with an external recorder like a lot of the other videos we've watched.
It doesn't look great. The R5 is surely better based on those videos. But it can be quite awkward to present the camera with a stable image for it to look for eyes when one is trying to record the LCD with a phone (ask me how I know).
Still, I shot Yellowlegs with the R5 that were doing similar things to that Stilt and the R5 holds the eye all the way down till the eye goes under the water line. Then with just half an eye coming back up out of the water the R5 will already be back showing the eye AF square. The A1 didn't even keep on the stilt's eye as it started to bend its neck down.
In some instances like the first part of the pelican sequence the Zone isn't even over the eye so of course the system won't recognize it. But later on in that video the Zone is over the eye and it is still having trouble.
Hopefully Sony will keep working on the BE-AF and provide FW updates considering how much this camera costs. Certainly BE-AF is not the reason to be buying this camera. The stacked sensor is the reason and everything it brings to the table over something like the R5.
The review mentioned that BEAF would disable itself whenever two eyes are identified (this factoid coming from Sony). Not sure whether this could explain the issue in the cluttered zoo, though. That pelican sure looked easy enough.
The review link I posted above is from a very popular Taiwan site. The reviewer himself is pretty well-known in that community and has tons of review about Sony gears. Although he is not a wildlife photographer but we get idea how A1 behaves on the day one. Some reviewer on FB states that A1 is about 80% of R5 on BE-AF. I would agree with such assessment based on all videos I have seen. Oh well, we are sure that Sony will improve it along the way.
Yes, I think the BEAF is fine but not the BE-D (bird eye detection), which still needs some work. Sony needs more and better training data. Their neural network architecture is pretty good since it works fine on humans, so it's just the learning set that must be improved for birds: they just need to train it with a lot more examples.
80% is not bad for a start, and from their interview, they seem to be aware that it's not optimal and will certainly address this in firmware updates.
The other part that's missing here for anyone who shoots video is the lack of animal and bird eye detect in video mode. Same thing goes for all the other Sony cameras that offer animal eye detect. I'd be happy if Sony were to state "we're working on it and expect it within "x" amount of time but all I hear are crickets.
Those are clips of BEAF shooting a TV. I'm not concerned yet. Is it perfect? No! When it misses, does it get a wingtip? NO its gets something close to the eye, head ....etc. I'm looking forward to getting mine and testing it out. I'll post some viewfinder and matching photos when I get one in my hands and head up to niagara falls where all the ducks and gulls are.... in 8-9 days :-)
buffalowolff wrote:
Those are clips of BEAF shooting a TV. I'm not concerned yet. Is it perfect? No! When it misses, does it get a wingtip? NO its gets something close to the eye, head ....etc. I'm looking forward to getting mine and testing it out. I'll post some viewfinder and matching photos when I get one in my hands and head up to niagara falls where all the ducks and gulls are.... in 8-9 days :-)
They are not shooting a TV. If you read the written review he talks about the 3 different places he visited to shoot those videos and test the different eye-af options. In the llama video (not linked to here but in the review) you can see all the other people around at the "zoo".
The written review is excellent. Dives into all sorts of things. Way more in depth than any other review I've seen to date.
arbitrage wrote:
They are not shooting a TV. If you read the written review he talks about the 3 different places he visited to shoot those videos and test the different eye-af options. In the llama video (not linked to here but in the review) you can see all the other people around at the "zoo".
The written review is excellent. Dives into all sorts of things. Way more in depth than any other review I've seen to date.
Teaches me to go from videos on my phone and pass judgement since they were recording the screen with a camera looked strange. Going through the whole thing translated its a pretty good review including their noise analysis.
octo wrote:
Yes, I think the BEAF is fine but not the BE-D (bird eye detection), which still needs some work. Sony needs more and better training data. Their neural network architecture is pretty good since it works fine on humans, so it's just the learning set that must be improved for birds: they just need to train it with a lot more examples.
80% is not bad for a start, and from their interview, they seem to be aware that it's not optimal and will certainly address this in firmware updates.
Lets hope so. Not holding my breath. I'm still waiting for a trivial star-eater firmware fix.
arbitrage wrote:
They are not shooting a TV. If you read the written review he talks about the 3 different places he visited to shoot those videos and test the different eye-af options. In the llama video (not linked to here but in the review) you can see all the other people around at the "zoo".
The written review is excellent. Dives into all sorts of things. Way more in depth than any other review I've seen to date.
Agree, well written and informative, esp. for people usually not looking into the more technically parts.
I would like to see many more reviews written in such a way.
Stoffer wrote:
... which reminds me: I wonder how this sensor will handle long exposures?
Ever since they added interval shooting into the menu system I have done my long exposures by stacking in PS. I rarely carry ND filters anymore. Enabling the camera to alter exposure (an option) in interval shooting also helps in fast changing light.
Meh. In his test the DR is similar, maybe a little worse on the a1 than the RIV and R5 . According to them: sharpness is similar across them, High ISO noise he says is similar, but he doesn't go through the ranges, just jumps up to 25600 which is useless anyway IMO. Interestingly enough they claim the stabilization is better on the a1 than all of the others, saying the R5 ties the a1, but his chart says the r5 ties the RIV in 2nd place... so who knows, but bodes well for the a1 I guess. anti-flicker works well.
).