How do I supply metadata when digitizing a print, slide, and negative library?

You have entered the twilight zone. Reality takes a back seat.

When you shoot digitally you get a bunch of information for free:

• A timestamp. You know to within a second (assuming your camera is set to the right time) when the picture was taken.
• The order that pictures were taken.
• Possibly GPS location.
• YOU took them (usually) so you know of the events covered.

When faced with digitizing someone else's collection, it's more difficult. A lot of this article is about recovering some of this data from the collection.

Another part are tips to maintain a back reference to the collection. Digital scanning has improved over the years, and will doubtless continue to improve. In many cases you want to be able to go back to the analog image. But to do that you have to know where to look.

I'm assuming that you are making a digital collection of images from your physical images -- prints, slides, negs.

I'm also assuming that you are doing the conversion with a digital camera. While a scanner can be used, the techniques for getting frame data are unknown to me. A digital camera can create the archive far faster than you can do the indexing. Scanners are much more problematic. Doing the multiple station setup to get frame data and image data with similar time stamps is far more difficult.

The following methodology assumes that you have a good digital camera 20 megapixels or more, along with a macro lens capable of focusing to 1:1.

First of all, you need to 'frame' your images. The following data will help establish context for your collection:

Time Stamp on slide Kodak, and at least some other processors would stamp the slide with month and year that the film was processed. Sometimes this was an ink stamp (usual on plastic mounts) or embossed into the cardboard on cardboard mounts.

Sidenote: The ink stamps are often alcohol soluble.

Ideas for recovering this information: If you use a digital camera for copying, set up 2 stations.

• Station 1 does the actual slide copy, with a macro lens focused on the slide. (BTW: Since slides are usually buckled, you may want to use stacked focusing for this.)
• Station 2 has enough back light to see the image, but is wide enough to see the entire slide. This station has an additional light from the side at a low glancing angle to bring out the date stamp.
• In addition slides have a frame number. This is in the same order as the film roll, but may not be the same as the actual number on the film. This can bite you when the better slides have been remounted in glass slide holders.
• Verify that the cameras for both stations are using the same time to the second. This way, you can sort by time and the station 2 image is adjacent to the station 1 image.
• If you get a good flow going, you can put both cameras on "take a shot every 20 seconds" and phase them 10 seconds apart. This may not be compatible with focus stacking unless you can do a focus stack burst every 20 seconds. Check into software for tethered shooting.

Other information from the mount

Sometimes people have written notes on the slide. Be prepared to photograph both sides. Also: a common trick for people who put together slide shows was to band the slides in order, then run a diagonal line with a felt marker down the block. This enabled quick restoration if a tray was fumbled. You can careful put a tick on the face of the slid corresponding to the middle of the edge mark. Sometimes the felt marker will bleed onto the edge enough to show up without a manual tick.

Frame numbers from negatives

The same trick works with negs: One shot or the frame, one shot backed off a bit to get the neg number. You may be able to do the second operation a strip at a time.

Don't discard the envelopes

If your collection arrives as a bunch of shoeboxes, examine the containers carefully. They will often have a date processed. This is best done by circling the date with felt marker, and if faint or small, copying it. Use a fine point permanent marker. Now take a picture on your station 2 setup.

Check the envelope for notes. More meta pictures.

Don't discard the receipts

A receipt will often tell you how many rolls were processed. This can help put the pieces together.

Construct a timeline for each box.

For prints and negs this will usually be a full date. For slides, only a month and year. Note that processing time is just that. Sometimes people would collect 3-4 rolls of film over the space of a year, then process them all at once. Or the last roll of the beach holiday wasn't used up, and shows up in January with the other Christmas pics.

So this is a tentative timeline. Each box gets a box number. At this point you won't know what order the boxes are. Each container (slide box, neg bundle, print bundle. gets a Box number and a date order. Where you have duplicate dates, arbitrarily add a 1 letter suffix. If the negs are with the prints, then their number is the same as the prints envelope with -neg as the suffix. this means you can have

• Box 6 1984-Jul-15a
• Box 6 1984-Jul-15a-neg
• Box 6 1984-Aug
• Box 6 1984
• Box 6 Unknown-a

Slides are often separated from their envelopes, so the only info you have is mmm-yy

Prints

You need a copy camera set up. Dual lights 45 degrees off axis. Ideal is a vacuum easel, but failing that, a slab of 1/4" polished plate glass (do NOT use float glass.) 1/4" thick gives you enough weight to hold the picture flat, and also for any dust specks on the top of the glass to be out of focus.

Note: You can try putting a weight on an envelope of photos for a few days, and the images may then lie flat enough. E.g. For my Tamron 90 mm used on a D7100 the depth of field at f/5.6 filling a frame with a 3x5 print is about 2 cm. I'm not going to worry about keeping the picture absolutely flat.

Ideal lights are LED lamps with a colour rendering index (CRI) of 90+ At this point I think you want daylight bulbs with a colour temperature of 5000-5500K. Buy all your bulbs at once, as there can be batch variations. You do not want one side of the stand to be relatively pink compared to the other. 2 or 4 lamps shining at an angle of about 45 degrees. Aim the brightest spot of the beam at the far edge.

If you can't get daylight bulbs, get tungsten temperature bulbs, but set the camera to that temp. Auto white balance didn't work for me. Correcting it in post did odd things to the contrast and saturation.

Prints often have a roll number and a date on the back. This roll number will be the same (usually) as the roll number on the end of the negative. This may give you a date for the negatives.

In this case station 2, is another camera with a wider field of view of the set up. Make a 'roll note' with the info from the envelope, and the common printed info on the back of the print. It is taped to your table near the pictures so you get both images. This camera only has to be good enough to be able to match the image to the roll info.

If you have negs with the prints, whatever info you can get off the prints transfer to the glassine sleeves the negs are in. If the negs are unsleeved, write a note that goes with the negs. When you process that set of negs, the first shot is whatever you have gleaned from the container info.

Computer processing

Each roll becomes an album or folder.

Add the meta data you salvaged from the packaging to the folder notes, and to the meta information for all the pictures in the folder.

If you know dates, use a date based folder name. If a folder seems to be all of an event, add the event string to the folder name. But the original Box #-date info from the timeline is also stored.

You will probably do negs in a separate run This is where the -neg comes in. If you can, process the negs into a positive image, and decide which you want to keep. Could be both. The neg station 2 shots will allow to to put the prints in order. Exiftool allows you to modify the dates in an image. If you do negs as a seperate run days or months later, you may want to subtract an appropriate interval to get them to the same time. Keeping a journal of your project with a list of what you did, when you started, when you finished may make this easier.

I thought of working tethered. I have an older macbook with an SD slot. The win would be not touching the camera, so less shake. Alas, the USB connection on that computer or the software makes for about a 15 second pause between taking the picture and showing it. This is about 5 seconds longer than it takes manually. Your mileage may vary if you have a faster computer.

The order is significant info. Several times when reconstructing events from a bundle of pix the order was helped me do so. If the fountain is frame 21, and frame 18 is the eiffel tower, then the fountain is the one in Paris.

Now, extract faces. Initially you won't have a clue, so every person initially is P:ZZ-NUMBER. (ZZ puts them all in the end of the list.)

Take your laptop with your pix when you see your parents. Now you can ID people. Once you know that P:ZZ-12243 is Uncle Jim, you can do globally change P:ZZ-12243 as "Jim Swenson -- according to Martha Swenson" (TEST that you can make global changes to metadata. You do NOT want to have to search each occurence separately.) You will get conflicts. You need to figure your own answer to this. My take would be to give that individual multiple labels. Eventually you may find that Martha's facial memory is terrible. (I'm nearly faceblind, I've discovered.)

Recap

• Make a timeline.
• Extract all container information you can.
• Start with a picture of your timeline.
• Start each roll with the roll's meta information. You can either photograph the container, or write it on a sheet of paper and photograph that.
• Scan the roll.
• Import the images into your library.
• Cull the blanks.
• Look at what you need to reshoot.
• Merge information into your archive.

Edit:

Storing the information:

Facial recognition requires support from your photo management program. This is usually created as some variation of a box with corners given in percent width and percent height. Note that certain editing operations: crops, rotations, flips, and perspective correction will wreck this and require that the image be reprocessed. Your program MUST have a way to extract this information in a known format.

If your facial recognition software isn't up to the task, creating a family facebook page and getting people to ID what they can may help. I've no idea how to get this info off of FB.

Keywords: Almost all data formats now support keywords. Worst case you create a database (Do NOT do this with a spreadsheet for any large collection.) The database contains the file name you gave it on your computer.

You can store any weird data (not usually stored with digital cameras) as keywords using sequences. E.g. FACE:21/42;15/70-"Mike Swenson according to Gloria Swenson" Frame:16 FMT:Neg Arch:Box=1984a-Jul-18a. Keyword field can have up to 64,000 bytes I think. You can also repurpose fields, although this can bite you.

Once you have it in a database, and have a program to pull the data out for a given image, you use exiftool in a script to pull the info from the database and write it to the image, or you can do your homework, and write it as sidecar files. (I don't know the format. I'd have to do homework.

TEST THIS. While ExifTool claims to be able to write metadata into most raw image formats, Phil doesn't have every camera. Duplicate a folder of images and then check for corruption.

Once you have all the station2 data extracted you can delete these images.

You can edit exif metadata with an editor, command line, or in photo software like Adobe Lightroom. You can adjust date, camera and lens info, and even add GPS coordinates.