x3 FFF- one other problem ...

Seems that a FF a few years from now will have to go against Fuji aggressive pricing of their 100 mp medium format cameras

with the 100s now 6 grand one of of 3 things could happen

1) Sigma would have to price the FF FOVEON at about 3 grand after 6 years of $$$$$$ development to attract potential Fuji MF customers

2) Fuj could counter with a 150 mp sensor

3) Fuji could create a dedicated Monochrome to steal the hi rez landscape hobbyists

4) advancements in AI software such as DXO " DEEP PRIME" are allowing the high iso images to reach astounding quality in even small micro 4/3 sensor cameras - this software wont work with foveon putting Sigma at an enormous disadvantage if the new sensor is a daytime only type camera in the same vein as the SD15

Mr Giggles wrote:

Seems that a FF a few years from now will have to go against Fuji aggressive pricing of their 100 mp medium format cameras

with the 100s now 6 grand one of of 3 things could happen

1) Sigma would have to price the FF FOVEON at about 3 grand after 6 years of $$$$$$ development to attract potential Fuji MF customers

2) Fuj could counter with a 150 mp sensor

3) Fuji could create a dedicated Monochrome to steal the hi rez landscape hobbyists

4) advancements in AI software such as DXO " DEEP PRIME" are allowing the high iso images to reach astounding quality in even small micro 4/3 sensor cameras - this software wont work with foveon putting Sigma at an enormous disadvantage if the new sensor is a daytime only type camera in the same vein as the SD15

Mr Giggles,

You've given me a giggle.

Why are you so concerned about what Fuji do or might do.

Sigma will do what Sigma does in their own good time & what they will do will stand on it's own. Looking forward to what ever they will produce.

S

Mr Giggles wrote:

just as an example .... here is one of my night images using a tiny micro 4/3 sensor camera processed with artificial intelligence program DXO DEEP PRIME

can you imagine if in 3 years the FF FOVEON camera cant shoot clean images over ISO 800 ?

notice that this was shot at iso 6400 - 1/500 second at night !

zoom in and pixel peep

What sensor is in your smart phone, that is tiny, m43 is a big sensor.

I really can imagine that the 'clean above iso 800 photos' are not possible at all with Foveon sensors, many scientists say so, and Sigma is working on it since more than a decade, with not much progress, so ... and there is a limit to ai.

You cannot compare the progress made in Bayer sensors to Foveon sensors.

Jozef.

Mr Giggles wrote:

just as an example .... here is one of my night images using a tiny micro 4/3 sensor camera processed with artificial intelligence program DXO DEEP PRIME

can you imagine if in 3 years the FF FOVEON camera cant shoot clean images over ISO 800 ?

notice that this was shot at iso 6400 - 1/500 second at night !

zoom in and pixel peep

The example is false because the main subjects are lit by bright artificial light, not the ambient light.

Mr Giggles wrote:

Seems that a FF a few years from now will have to go against Fuji aggressive pricing of their 100 mp medium format cameras

with the 100s now 6 grand one of of 3 things could happen

1) Sigma would have to price the FF FOVEON at about 3 grand after 6 years of $$$$$$ development to attract potential Fuji MF customers

2) Fuj could counter with a 150 mp sensor

3) Fuji could create a dedicated Monochrome to steal the hi rez landscape hobbyists

4) advancements in AI software such as DXO " DEEP PRIME" are allowing the high iso images to reach astounding quality in even small micro 4/3 sensor cameras - this software wont work with foveon putting Sigma at an enormous disadvantage if the new sensor is a daytime only type camera in the same vein as the SD15

The Bayer sensors and Foveon sensors (cameras) are absolutely different in "ideological" and technical aspects and cannot be compared in any way.

The DxO Deep prime is not the real Ai because the Ai methods involve neural training on tens of thousands different images (obtained under different conditions) from different cameras. Does DxO have a collection of these test shots? I don't think so... This is only the successful modification of prime hi-iso noise reduction method which only works for Bayer sensors (raw-files).

The computation photography/software (Photo Acute or etc.) and pixel-shift can give better results than DxO Deep Prime, but implies the use of stacking and static objects/subjects.

How sigma will solve this Hi ISO issue is still unknown. But I will wait and will buy the future FF-foveon camera )

dpreviewblog wrote:

Mr Giggles wrote:

Seems that a FF a few years from now will have to go against Fuji aggressive pricing of their 100 mp medium format cameras

with the 100s now 6 grand one of of 3 things could happen

1) Sigma would have to price the FF FOVEON at about 3 grand after 6 years of $$$$$$ development to attract potential Fuji MF customers

2) Fuj could counter with a 150 mp sensor

3) Fuji could create a dedicated Monochrome to steal the hi rez landscape hobbyists

4) advancements in AI software such as DXO " DEEP PRIME" are allowing the high iso images to reach astounding quality in even small micro 4/3 sensor cameras - this software wont work with foveon putting Sigma at an enormous disadvantage if the new sensor is a daytime only type camera in the same vein as the SD15

The Bayer sensors and Foveon sensors are absolutely different in "ideological" and technical aspects and cannot be compared in any way.

The DxO Deep prime is not the real Ai because the Ai methods involve neural training on tens of thousands different images (obtained under different conditions) from different cameras. Does DxO have a collection of these test shots? I don't think so... This is only the successful modification of prime hi-iso noise reduction method which only works for Bayer sensors (raw-files).

The computation photography/software (Photo Acute or etc.) and pixel-shift can give better results than DxO Deep Prime, but implies the use of stacking and static objects/subjects.

How sigma will solve this Hi ISO issue is still unknown. But I will wait and will buy the future FF-foveon camera )

I don't see why Sigma can't offer a high-ISO line based on the fp, and a lower-ISO high definition line for careful use in good light based on some kind of Foveon sensors. If the Foveon can be made to give really low noise at ISO 400, it will cover everything that used to be done with medium-format cameras.

The fp type then covers everything that used to be done with 35mm film cameras, plus video.

Pentax, for example, sold both 35mm and medium format cameras.

AI may become useful for post-processing but I don't see it being useful in cameras. IMO it is greatly over-rated at present.

Don Cox

D Cox wrote:

dpreviewblog wrote:

Mr Giggles wrote:

....

4) advancements in AI software such as DXO " DEEP PRIME" are allowing the high iso images to reach astounding quality in even small micro 4/3 sensor cameras - this software wont work with foveon putting Sigma at an enormous disadvantage if the new sensor is a daytime only type camera in the same vein as the SD15

The Bayer sensors and Foveon sensors are absolutely different in "ideological" and technical aspects and cannot be compared in any way.

The DxO Deep prime is not the real Ai because the Ai methods involve neural training on tens of thousands different images (obtained under different conditions) from different cameras. Does DxO have a collection of these test shots? I don't think so... This is only the successful modification of prime hi-iso noise reduction method which only works for Bayer sensors (raw-files).

The computation photography/software (Photo Acute or etc.) and pixel-shift can give better results than DxO Deep Prime, but implies the use of stacking and static objects/subjects.

How sigma will solve this Hi ISO issue is still unknown. But I will wait and will buy the future FF-foveon camera )

I don't see why Sigma can't offer a high-ISO line based on the fp, and a lower-ISO high definition line for careful use in good light based on some kind of Foveon sensors. If the Foveon can be made to give really low noise at ISO 400, it will cover everything that used to be done with medium-format cameras.

The fp type then covers everything that used to be done with 35mm film cameras, plus video.

Pentax, for example, sold both 35mm and medium format cameras.

AI may become useful for post-processing but I don't see it being useful in cameras. IMO it is greatly over-rated at present.

Don Cox

+100

IMO may be the "unlimited" in-camera stacking (as in Pentax K-1 / K-1 MII) is one of the future method for Sigma FFF to reduce noise. And the recent SFD mode is the beginning of future Sigma attempts to control/reduce noise via computational methods? )

Mr Giggles wrote:

xpatUSA wrote:

Mr Giggles wrote:

just as an example .... here is one of my night images using a tiny micro 4/3 sensor camera processed with artificial intelligence program DXO DEEP PRIME

can you imagine if in 3 years the FF FOVEON camera cant shoot clean images over ISO 800 ?

notice that this was shot at iso 6400 - 1/500 second at night !

zoom in and pixel peep

AI DXO "deep prime" is a huge breakthrough - the proof is in the pudding

and Sigma will need not just a new FF camera - they will need a breakthrough raw processor too

The example is false because the main subjects are lit by bright artificial light, not the ambient light.

I have many other examples

all of which are of subjects lit by bright artificial lighting.

My point is that just about any Sigma camera could have captured those scenes well enough.

On the other hand, my DC-G9 would most certainly beat my SD9 on a moonlit night with no artificial lighting at all.

Mr Giggles wrote:

xpatUSA wrote:

Mr Giggles wrote:

xpatUSA wrote:

Mr Giggles wrote:

just as an example .... here is one of my night images using a tiny micro 4/3 sensor camera processed with artificial intelligence program DXO DEEP PRIME

can you imagine if in 3 years the FF FOVEON camera cant shoot clean images over ISO 800 ?

notice that this was shot at iso 6400 - 1/500 second at night !

zoom in and pixel peep

AI DXO "deep prime" is a huge breakthrough - the proof is in the pudding

and Sigma will need not just a new FF camera - they will need a breakthrough raw processor too

The example is false because the main subjects are lit by bright artificial light, not the ambient light.

I have many other examples

all of which are of subjects lit by bright artificial lighting.

My point is that just about any Sigma camera could have captured those scenes well enough.

On the other hand, my DC-G9 would most certainly beat my SD9 on a moonlit night with no artificial lighting at all.

I had a SD9 myself ... no way it takes these images .... look at the dark shadow areas - not the brighter areas under the lights

No need to - we should all be well aware that scenes with such contrast have light areas and dark areas.

We may be talking at cross-purposes ... If I were to shoot that gas station at that time with my SD9, I wouldn't care at all about the dark shadows.

10EV natural Lighting Value - captured well enough for me ...

Mr Giggles wrote:

You cannot compare the progress made in Bayer sensors to Foveon sensors.

Jozef.

That's the problem - Foveon is now up against not only bigger - cheaper and better bayer cameras - its up against much better RAW processors and artificial intelligence

its a scary situation for Sigma

It isn't

Sigma is a lens maker, cameras are just a side project from the Sigma ceo Yamaki, and he has always lost money on his camera projects, that is the way I see it.

It is much better for Sigma's financial situation if they leave the camera business for what it is, and concentrate solely on lens manufacturing. My opinion.

Jozef.

Mr. Giggles,

These shots are very very impressive IMHO.  This is good news for Sigma no?  Topaz and Capture one work on Tiffs and Jpgs no?

One question, I would like to see the original to do a comparison, but these look outstanding to my eye.

dpreviewblog wrote:

The DxO Deep prime is not the real Ai because the Ai methods involve neural training on tens of thousands different images (obtained under different conditions) from different cameras. Does DxO have a collection of these test shots? I don't think so... This is only the successful modification of prime hi-iso noise reduction method which only works for Bayer sensors (raw-files).

This is what they say:

https://www.dxo.com/dxo-photolab/deepprime/

Thanks to our vast portfolio and the billions of carefully selected input and output samples uploaded onto DxO DeepPRIME, our AI technology has ‘learned’ the two operations fundamental to editing any RAW image file – demosaicing and denoising.

bizi clop wrote:

dpreviewblog wrote:

The DxO Deep prime is not the real Ai because the Ai methods involve neural training on tens of thousands different images (obtained under different conditions) from different cameras. Does DxO have a collection of these test shots? I don't think so... This is only the successful modification of prime hi-iso noise reduction method which only works for Bayer sensors (raw-files).

This is what they say:

https://www.dxo.com/dxo-photolab/deepprime/

Thanks to our vast portfolio and the billions of carefully selected input and output samples uploaded onto DxO DeepPRIME, our AI technology has ‘learned’ the two operations fundamental to editing any RAW image file – demosaicing and denoising.

Thanks a lot for the link, but IMO - you know, you can't trust everything you read on... ) Better to discuss this in another forum thread related to software, etc. - Photographic Science and Technology - https://www.dpreview.com/forums/1061 , moreover the Bayer demosaicing has nothing to do with Foveon