Jump to content
The Education Forum

Recommended Posts

2 hours ago, Mathias Baumann said:

So I just had an e-mail exchange with Dr Rudin. As I suspected her software is not suitable for identifying Prayer Man. In fact, she wrote that

I guess that means we'll never know who Prayer Man is. There's just not enough information in the picture.

I guess it only means that Dr Rudin's method is not suitable for disclosing with certainty the true appearance of someone's face if the image is blurred. It would still be interesting to see what face would come out from their AI analysis if the face of Prayer Man would be used as input.

Link to post
Share on other sites
  • Replies 248
  • Created
  • Last Reply

Top Posters In This Topic

16 hours ago, Sandy Larsen said:

 

I would have to know Duncan McRae to make judgment on him. Chris Davidson is being fooled by what appears to be a lovely woman's face. I happen to agree with him that -- after his filtering of the image -- it appears there is a lovely woman’s face in it. But I know intellectually that we are both seeing a "shape in the clouds." You can't gain true image details through the use of generic Photoshop filters. As I said, you need to use custom filters.
At least that is the case for anti-blurring. If, on the other hand, you have a grainy image, a general low-pass filter can be used to eliminate some of that noise.
Now, there are some who see a woman even without the use of filters. I have no idea how they see a woman because PM looks obviously like a man to me. I suppose it could be that they want to see a woman because they are trying to figure out where a missing woman is, like Sarah Stanton. (I figure she must be standing behind someone. In fact, Andrej Stancak has located what appears to be part of a head just behind Frazier's head.)
 

 

 

In digital signal processing, an image is treated as a two-dimensional signal after it has been digitized (converted to a string of numbers or digits). A digitized sound or radio wave are treated as one-dimensional signals.

 

thank you.

Link to post
Share on other sites

Chris,

I have looked at your .gif and have taken out certain frames that are in a rough sequence:

pm-davidson-camera-flash.jpg

It looks like PM has a camera that has flashed.  I would like to get your opinion on that.  Is there something going on here that explains the circles of light as something other than a camera?

I have thought for years that PM has a camera in his hands rather than a coke or some other object.

OBTW,

It does look like a woman's face anthropomorphically speaking:

pm-woman-s-face.jpg  

Edited by John Butler
Link to post
Share on other sites
3 hours ago, Chris Davidson said:
17 hours ago, Sandy Larsen said:

I would have to know Duncan McRae to make judgment on him. Chris Davidson is being fooled by what appears to be a lovely woman's face. I happen to agree with him that -- after his filtering of the image -- it appears there is a lovely woman’s face in it.

 

Expand  

I wasn’t fooled at all. That is your mis-characterization of what I said after I ran the “shadow contrast and tonal curve” filters and stated “It looks like a woman to me.”

I did not, nor have I ever declared it was a woman. Big difference.

 

I didn't say that you declared it was a woman. I said that you thought it was a woman. And at that time you did. I know because I kept asking what you thought and you kept skirting the issue. Instead you told me that your wife thought it was a woman.

Had you not thought it was a woman, you'd have said so.

The only mischaracterization I may be guilty of is that I used the present tense when I commented on what you thought. I should have used past tense because you may have changed your mind since then.

 

Quote

There were two frames in which the filter was applied one was more blurry than the other.
The more blurred frame(after applying the filter) was the one in which I made the comment about looking like a woman.
The other frame (also with the filter applied) was the one that Duncan enlarged, which is the one you are referring to and I have never said that frame looked like a woman regardless of whether they depicted the same person.

 

I referred to both Duncan's "Frankenstein" woman's image and to your "lovely face" woman's image in my comments.

 

Edited by Sandy Larsen
Link to post
Share on other sites
2 hours ago, John Butler said:

pm-davidson-camera-flash.jpg

It looks like PM has a camera that has flashed.  I would like to get your opinion on that.  Is there something going on here that explains the circles of light as something other than a camera?

 

John,

The outer lighter ring which extends around the head/arms of PrayerPerson is the area I selected using Photoshop to lighten the image. The far left image(without the lightened circle) is the original frame before the enhancements.

There are three separate frames in that previous gif, the first frame in each is the original unenhanced frame.

The white object being held is anyone's guess. Camera, coffee mug etc, etc

 

Link to post
Share on other sites

I would like to comment on the alleged female seen in the Wiegman still and hope not to add to the confusion. The alleged image of a female in Wiegman was brought here by one former member of the Forum back in 2016. The thread is named "Prayer Person - Man or Woman? and the thread has been moved outside the Debate folder to JFK Research folder where it can still be visited.

After quite a number of repeated requests, the image of alleged female was posted and it was the small image in the right-most panel of the composite picture below. After further requests to show the original, unprocessed image from which this trimmed image was produced, Chris Davidson kindly posted the Wiegman still on which he did his enhancements.  This original image is the left-most panel in the composite below. I would like to ask everyone to inspect this original image and decide if it is possible to see more than a very faint contour of a person standing at Prayer Man spot and whether there can be any confidence in saying that that person was a male or female.

The enhancements which Chris made on the original still, as I understand it, were only in form of adding some brightness and contrast. I have replicated these steps and found the same figure  as Chris did (middle panel in the composite below).  I see distorted information in the head and neck region due to low signal in the original image (left). The enhancement algorithms, unfortunately, do not differentiate between real information and noise and make both type of information prominent. Please see the misalignment of the upper and lower part of the head which is typical in noisy images having also low resolution.

There was someone else who cut the part of Prayer Man's head from Chris'es enhancement, namely the lower face, neck and upper chest, and made it a new picture and claimed it was a female. The image was "enhanced" further, rotated by 2 degrees and trimmed. However, the composite image clearly shows that that image 1) originates from an area of Wiegman still which contains very little signal and any enhancements, therefore, lead to spurious results, 2) it was trimmed not to show the upper part of Prayer Man's head. The image of alleged lady which some claim to be Sarah Stanton is a photographic misinterpretation resulting from the lack of understanding of the original picture or even an intentional misleading of research community.

 

 

pm_wieglady-1.jpg 

This next picture shows, side-by-side, the figure of Prayer Man in Darnell and in the enhanced Chris Davidson's still from Wiegman film. Please note that the true contour of Prayer Man figure in Wiegman also includes the top part of the head above the line of trimmed image of alleged female face.

 

pm_wiegdarnell.jpg

 

 

Edited by Andrej Stancak
Link to post
Share on other sites
11 hours ago, Chris Davidson said:

The white object being held is anyone's guess. Camera, coffee mug etc, etc

Thanks Chris,

For another neat and interesting gif demonstration.  I think I answered my question a couple of hours later.  As much as I wanted that to be camera flash bulb, one has to go with reality.  A camera flash would be more diffuse and scattered while that lightened circle had to be something else.  

As far as the object in PM's hands, your correct.  It is far to vague to actually make a call.  It could be a camera, coffee cup, a coke, etc.  

I'm left with it looking like a camera due to PM holding it close to his face with two hands.  If it was a camera I could link it to the Prayer Man/Oswald figure on Elm Street by the western end of the TSBD taking photos as the p. limo passed by that one can see in the John Martin film.

If that linkage was made it would really screw things up.  It would give you 3 Oswalds.  And, nobody, including myself, would believe that.  You would have the PM figure on Elm St. taking photos as the p. limo goes by and then run to the door way of the TSBD before Marion Baker and become the PM in Darnell/Couch, a Doorway figure in Altgens 6 made up of somebody (Oswald?) and a Billy Lovelady composite who maybe goes up to the 2nd floor.  Then we have the Oswald on the 6th floor who walks down the steps after the assassination and meets Truly and Baker at the 3rd or 4th floor.

Way to many Oswalds when only two can be seen at the TSBD when one leaves in a Rambler and another takes a bus.  One could cut out an Oswald if you said the the figure on the 3rd or 4th floor was someone else.  And, that the 6th floor Oswald really came down to the doorway prior to the assassination and was out front with Bill Shelley.  And, from there walk up to the 2nd floor for a coke.

Way to much speculation going on here.   

  

Link to post
Share on other sites
21 hours ago, Andrej Stancak said:

I guess it only means that Dr Rudin's method is not suitable for disclosing with certainty the true appearance of someone's face if the image is blurred. It would still be interesting to see what face would come out from their AI analysis if the face of Prayer Man would be used as input.

Andrej,

I'll see if I can the Python software running on my computer. But I suspect I'll need a more powerful GPU to perform this kind of calculation. Unfortunately I only have a run-of-the-mill desktop PC.

Link to post
Share on other sites

Mathias:

please let me know if I can be of any assistance. Once you upload their libraries into Python and run the code with their sample image, I would prepare a suitable image of Prayer Man's face. The Darnell still  we all use in our posts, although giving some interesting details, has the nose and eyes areas burned out, and no AI could reconstruct anything unless there is at least some signal in the image.

 

Link to post
Share on other sites
9 hours ago, Andrej Stancak said:

I would like to comment on the alleged female seen in the Wiegman still and hope not to add to the confusion. The alleged image of a female in Wiegman was brought here by one former member of the Forum back in 2016. The thread is named "Prayer Person - Man or Woman? and the thread has been moved outside the Debate folder to JFK Research folder where it can still be visited.

After quite a number of repeated requests, the image of alleged female was posted and it was the small image in the right-most panel of the composite picture below. After further requests to show the original, unprocessed image from which this trimmed image was produced, Chris Davidson kindly posted the Wiegman still on which he did his enhancements.  This original image is the left-most panel in the composite below. I would like to ask everyone to inspect this original image and decide if it is possible to see more than a very faint contour of a person standing at Prayer Man spot and whether there can be any confidence in saying that that person was a male or female.

The enhancements which Chris made on the original still, as I understand it, were only in form of adding some brightness and contrast. I have replicated these steps and found the same figure  as Chris did (middle panel in the composite below).  I see distorted information in the head and neck region due to low signal in the original image (left). The enhancement algorithms, unfortunately, do not differentiate between real information and noise and make both type of information prominent. Please see the misalignment of the upper and lower part of the head which is typical in noisy images having also low resolution.

There was someone else who cut the part of Prayer Man's head from Chris'es enhancement, namely the lower face, neck and upper chest, and made it a new picture and claimed it was a female. The image was "enhanced" further, rotated by 2 degrees and trimmed. However, the composite image clearly shows that that image 1) originates from an area of Wiegman still which contains very little signal and any enhancements, therefore, lead to spurious results, 2) it was trimmed not to show the upper part of Prayer Man's head. The image of alleged lady which some claim to be Sarah Stanton is a photographic misinterpretation resulting from the lack of understanding of the original picture or even an intentional misleading of research community.

 

 

pm_wieglady-1.jpg 

This next picture shows, side-by-side, the figure of Prayer Man in Darnell and in the enhanced Chris Davidson's still from Wiegman film. Please note that the true contour of Prayer Man figure in Wiegman also includes the top part of the head above the line of trimmed image of alleged female face.

 

pm_wiegdarnell.jpg

 

 

I showed Prayer Man to a friend of mine, who is an expert in computer graphics. He says the image quality and resolution are way too poor to tell anything with reasonable certainty. The reason is it is not possible to know which pixels represent real information and which just digital artifacts.

And I think this is further confirmed by Dr Rudin's remarkable observation. What you posted here is a good example. The pixels that make up Prayer Man's face can correspond to a great variety of different faces. We'll never be able to tell with certainty if Prayer Man is man or woman, let alone what shirt he/she is wearing or what he/she is holding in their hands. The information in the picture is just insufficient. 

Link to post
Share on other sites
1 minute ago, Andrej Stancak said:

Mathias:

please let me know if I can be of any assistance. Once you upload their libraries into Python and run the code with their sample image, I would prepare a suitable image of Prayer Man's face. The Darnell still  we all use in our posts, although giving some interesting details, has the nose and eyes areas burned out, and no AI could reconstruct anything unless there is at least some signal in the image.

 

Can you point me to a source where I can download the entire film in high resolution?

Link to post
Share on other sites
14 minutes ago, Mathias Baumann said:

Can you point me to a source where I can download the entire film in high resolution?

Mathias:

there is no Darnell film of high resolution available to the research community, unfortunately. I work with a set of images which I believe were exported by Robin Unger several years ago. I can upload the folder onto Dropbox if you wish.

The image I believe contains facial features that could possibly be enhanced using AI is this:

frame_3520_single-e1561315080133.jpg

 

Here is the same image (the left-most one) with progressively more intense overlays of Lee Oswald's head. Amazingly, the orientations of Prayer Man and Lee Oswald's head in both images are practically identical.

In the original image, the small bright spot below the chin is most likely Adam's apple. Lee had an elevated nasolabial region, rising toward the upper lip. I guess this can be seen in the original picture. If there is any picture offering a chance of revealing Prayer Man's facial features, it would be this one.

bigoverlay-1.jpg 

 

 

Edited by Andrej Stancak
Link to post
Share on other sites
3 minutes ago, Andrej Stancak said:

Mathias:

there is no Darnell film of high resolution available to the research community, unfortunately. I work with a set of images which I believe were exported by Robin Unger several years ago. I can upload the folder onto Dropbox if you wish.

The image I believe contains facial features that could be possibly enhanced is this:

frame_3520_single-e1561315080133.jpg

 

Here is the same image (the left-most one) with progressively more intense overlays of Lee Oswald's head. Amazingly, the orientations of Prayer Man and Lee Oswald's head in both images are practically identical.

In the original image, the small bright spot below the chin is most likely Adam's apple. Lee had an elevated nasolabial region, rising toward the upper lip. I guess this can be seen in the original picture. If there is any picture offering a chance of revealing Prayer Man's facial features, it would be this one.

bigoverlay-1.jpg 

 

 

Andrej,

I think what Dr Rudin was trying to say is this: the output of her software is determined by the input. So if you feed it a picture of Oswald you'll get Oswald as an output. If you feed it Sarah Stanton you'll get Sarah Stanton. You can't use it to enhance the quality of the image. It's only useful if you already know who's in the picture and if you have high-quality photos of that person that you can feed into the software.

Link to post
Share on other sites
Just now, Mathias Baumann said:

Andrej,

I think what Dr Rudin was trying to say is this: the output of her software is determined by the input. So if you feed it a picture of Oswald you'll get Oswald as an output. If you feed it Sarah Stanton you'll get Sarah Stanton. You can't use it to enhance the quality of the image. It's only useful if you already know who's in the picture and if you have high-quality photos of that person that you can feed into the software.

Mathias:

their work would make no sense if you only receive the same image which was used to scramble it and then reconstruct. However, I understand that authors want to be on a safe side and do not want someone falsely identifying a person in blurred images and possibly arrive at wrong conclusions. Anyway, I leave it with you whether you want to follow this line, and would understand fully if you decide not to.

Link to post
Share on other sites
14 hours ago, Andrej Stancak said:

Mathias:

their work would make no sense if you only receive the same image which was used to scramble it and then reconstruct. However, I understand that authors want to be on a safe side and do not want someone falsely identifying a person in blurred images and possibly arrive at wrong conclusions. Anyway, I leave it with you whether you want to follow this line, and would understand fully if you decide not to.

Andrej,

You could be right about that. Maybe PULSE could be used to identify Prayer Man by trial-and-error, that means by feeding it different images of different people and see which output looks most realistic. I don't know.

But I see a technical problem here:

Quote

Currently the code only works with CUDA installed (and therefore requires an appropriate GPU)

https://github.com/adamian98/pulse

So it appears you need some special hardware. CUDA was developed by Nvidia. I have a AMD RADEON HD 6450 installed in my computer. I don't think it'll support CUDA, but I'll give it a try anyway.

 

Link to post
Share on other sites

Please sign in to comment

You will be able to leave a comment after signing in



Sign In Now
×
×
  • Create New...