An app that lets you calculate the focal length used for (almost) any photo

We've often wondered what focal length a photographer used to take this or that photo. Some files have metadata, while some don't. For those that don't, there's an app called fSpy that lets you approximate the focal length used:

 

Link: fSpy

 

The catch is that the photo must contain straight lines in the scene. It's intended for use within 3D apps such as Blender (I saw a demonstration of this today). But it's a fairly handy tool just on its own.

 

I am pretty sure, I can see difference between 35, 50 and 85mm without computer.

How about between 28 and 35, or 35 and 40?

I think you're talking about uncropped photos.

Naturlich! If you crop a photo taken with a 28 mm lens to the FOV of a 35 mm lens, the perspective will be that of a 35 mm lens.

And the depth of field also?

I think you're talking about uncropped photos.

Cropping is not relevant, because the only thing that matters is AOV. Focal length is then derived from that, once you specify an arbitrary format.

And the depth of field also?

I'm fairly sure you can't derive that from a 2D image. That would probably require a stereo camera. Either that, or you would need to look at the metadata if it was available, in which case you would have your answer.

Cropping is not relevant, because the only thing that matters is AOV.

 

Wrong! Angle of view is changed by cropping.

Wrong! Angle of view is changed by cropping.

No, I am correct. Once you have a photo - cropped or not - that has straight lines, you can load it into this app and it will estimate the AOV view is for that image. Not the 'original' image. The image at hand. I know it's a little confusing, but just think about the image you actually have, and what lens you need to match to a given format to achieve that AOV.

 

Edit: Note that the app itself needs more data to be precise, which the user usually won't have access to. Ideally you should be able to estimate the physical dimensions of objects in the frame, which would help a lot.

Edited May 23, 2019 by Karim Ghantous

Wrong! Angle of view is changed by cropping.

Hector is correct. The program might estimate the effective angle of view from a cropped image, but it cannot analyze data which has been removed. You don't know what you don't know.

 

Perspective (convergence, etc) is determined by the distance from the subject(s), not the lens. A wider angle lens simply takes in more of the scene. This brings into question whether data from the image alone can be used to derive the focal length used. There must at least be some measure of distance, or more directly, metadata. I am not inclined to defend the indefensible.

(This all came up because I'm too old and cheap to buy a new Leica CL with a 35mm Summicron and instead bought a used Nikon Coolpix A with a fixed 28mm lens. Now with a good pair of scissors I'm good to go.) Thanks

Hector is correct. The program might estimate the effective angle of view from a cropped image, but it cannot analyze data which has been removed. You don't know what you don't know.

All photos are crops. A 35mm frame taken with an 80mm lens is a crop of a Hasselblad frame. Etc. But the photo you have is the photo you have. What the app doesn't let you do is measure distance or size, which would be a great help. Still, it works for its intended purpose, which I witnessed first hand, as I explained in the OP.

 

I should point out that this tool is intended for images that have no metadata or exposure notes. It is indeed better than nothing.

 

So that I am not misunderstood: the AOV of the image you have is what matters. Whether or not the image is cropped does not matter.

I recommend some outside reading. "The Camera" by Ansel Adams would be a good start. This book has an excellent discourse on optics and perspective.

I should point out that this tool is intended for images that have no metadata or exposure notes. It is indeed better than nothing.

The most expensive tool is one which does not accomplish what you need it to do. Second most expensive is a tool that does a job you do not need to do.

 

The question you must ask yourself is "Which lens should I use for this photo?" Where composition is the goal, this usually is predicated on the right position to use the compositional elements in the scene, followed by the focal length to capture them in the image. There are other considerations too. You can't get any closer if there is a fence between you and the elephants, or a 3000 foot drop at the Grand Canyon. Sometimes it is not convenient to change, or even carry more than one lens, street photography and urban touring on foot among them. As often as not, opportunity transcends composition.

 

I don't walk around the Chicago Botanic Garden with a 400 mm lens in tow, now through Millennium Park downtown. In fact using anything longer than 50 mm or shorter than 24 mm is reserved for special situations. Most of all, though, I can't imagine consulting an application to tell me what I should use, or what someone else used.

 

The possible exception is for astrophotography, where it is sometimes hard to judge the angle of view needed without consulting a reference. Orion, for example, is huge by eye, but fits nicely in the frame of an 85 mm lens. As a rule, only the brightest stars can be seen in the viewfinder.

The most expensive tool is one which does not accomplish what you need it to do. Second most expensive is a tool that does a job you do not need to do.

True! Tools matter.

 

Most of all, though, I can't imagine consulting an application to tell me what I should use, or what someone else used.

If you're curious, or if you're a 3D modeller (which is the target user for fSpy), yes, that stuff matters. A historian might want to know the probability of photographer X using a given focal length. If exposure notes apparently do not exist, he might want to examine the photo with an app like fSpy. Does it matter to the average person? I don't think so. But you would be surprised at what data is important and what isn't (e.g. I'm watching the case of Maddie McCann, and there is a lot of data involved there, plus the data that we are not privy to; there were also accusations of manipulated photos).

 

BTW the person who showed me this app in action was a 3D modeller who until recently worked on big studio projects. I would not have heard of this app otherwise. It's not a Holy Grail. But I figured that people would be moderately interested.

In contrary to others. I think the app is useful. I don't know how well it works or how to use it really.

Here are two photos that I just took. One is shot at 24mm and cropped, and the other is shot while standing in the same spot at approximately 85mm and uncropped. (Canon 5D IV, Canon 24-105mm L zoom, ISO 640, 1/60s, f:16, handheld, no lens correction). Can you tell the difference? Can fSpy tell the difference without looking at metadata? I don't think so. If you know the flare and resolution characteristics of the lens at the two focal lengths, perhaps you can guess the difference, but going by the perspective geometry of the two images, they should be identical.

Edited May 25, 2019 by Glenn McCreery

Why on earth would you really need the app in the first place?

Just for the heck of it, I threw several film scans at it which have no metadata.

 

Several were scans of slides in paper mounts, which I usually crop a decent amount so as to get a truly rectangular frame. I compared the slide scans with scans of negatives, which I do generally scan at 100%.

 

For the sake of comparison, I used photos that I remembered taking with the Canon nFD 20mm f/2.8. The negatives were reported as 20.5mm-21mm, which I'd call a pretty darn close estimate. The slides came back as 22-23mm.

 

As a further experiment, I tried inputting a custom "sensor size" and estimated that with my typical crop of a paper mount, I lose 1mm on each edge, or 2mm per dimension. Thus, I put the size in as 22mmx34mm, and when I did that the result came back as right at 20mm.

 

I also tried it to see what it would make of medium format scans, but it wouldn't me input 56mmx56mm or any other square aspect ratio(IIRC, 56x56 is roughly the actual size of a Hasselblad frame-some that I tried were taken with a Hasselblad, and some with a Bronica, but I'm assuming the Bronica is close enough to that).

but going by the perspective geometry of the two images, they should be identical.

They should be - and they are!

Why on earth would you really need the app in the first place?

If you use 3D software for any reason (e.g. Blender, Maya, Lightwave). Or, if you're just curious.

Just for the heck of it, I threw several film scans at it which have no metadata.

I didn't think it would be that accurate, but that's very good indeed. I guess you could use a 4:3 aspect as a close approximation, and translate from there. I am a little surprised that they don't have a square aspect to chose from.

 

"I can't believe it, you've actually found a practical use for geometry!" - Bart Simpson