Packard Forum Index » V-8 (1955-56)

All the Material cards should now be in the photo archive.

Great, in the meantime, I've been working on some material to make the audit process go more quickly and without introducing any errors.

Can you provide me with spreadsheet or even a simple text file of the gallery titles and their corresponding image filenames?

EDIT (2/26/2011, 12:06PM) - Cancel the above request. I ran the site search feature on the Packard Photo Archive this AM, using the keyword "Set", to pick up the material card titles. Then, rather than the image ID #s, what I really needed was the gallery entry ID #s, which were embedded in the link. Working one search page at a time, I highlighted the titles, and copied them to a text file, manually adding the gallery ID, and deleted the unecessary verbage. I'll report back here later, when I have some results.

A little update on my progress - boring stuff, but you can't build a great house without a solid foundation.

After skimming titles from the Search results to begin filling in a spreadsheet for analysis, I initially came up with only 362 entries (not counting one non-applicable photo that happened to contain the string 'set'). Oddly, that figure falls short of the 375 pages in the original PDF collection.

Yet, while transcribing the gallery ID numbers from the search results, page-by-page, I noticed some gaps in their sequence. So, I made a point to note the skipped ID numbers, and after splicing them into the browser URL, one-by-one, I picked up another 30 titles. Turns out that each was missing the term "Set". However, this drove the total over the top to 392, which was puzzling.

So, I copied the titles in my spreadsheet to a second column, stripped each one down to just the part number (P/N), and then resorted the table in P/N order. That revealed over two dozen genuinely duplicated entries. I don't know whether that was the result of overlapping work from two people on one project or the reported problem with an FTP upload session, but we'll work that out later. I left those entires in place in my spreadsheet, but deducting the count of duplicate entries dropped the total down below 375.

Next, I compared my spreadsheet to a list that I had composed a few weeks ago, identifying part numbers (P/Ns) by page numbers in the PDF collection, which turned up another 11 missing P/Ns. I went back to the Photo Archive and searched by part number to pick-up those titles. Again, each was missing the word "set", but those gallery ID numbers had not been apparent from my original Search results. Understand that while that numbering would be continuous within any given batch upload, the Search results were in alphabetical order, which chopped things up. I might have found these additional gaps if I had sorted my spreadsheet by gallery ID number, but hindsight is always 20/20.

Further inspection revealed a few more duplicates and even one set of triplicates, and I found a couple of digits transposed in two P/Ns. I also found a few missing images, but two of them appear to be redundant in the original PDF set. I still can't get the bottom line to balance out exactly, but it looks like we're within +/-1.

This week, I'll begin auditing the titles for any typos or inconsistencies and look into the parts books for expanded model coverage. However, working line-by-line, that's gonna take some time.

Brian, my compliments to you for undertaking the database. Also to patgreen and Stephen that did the scanning and labeling in the first place and to Roscoe for the donation. Just thinking about the detail entries and tedium that must be (and have been) involved is enough to give one a headache. Kudos to you all.

It will be a great resource for someone down the line.

I ended up redoing all of patgreen's work to keep a continuity (the tighter crop) after he had drop out because of his eye surgery.

I had found at least one entry that had two separate entries on its p/n tag and two separate entries from the parts catalog book I downloaded from this site. So I entered two separate files under that p/n. Also found different part numbers for different sets or trim codes covering several years. I reported some of the anomalies to Kev. Don't know if that is part of what you are having trouble with in the descriptions. Also the full tag entry was put into the file title to give the widest area of cross-reference for someone searching for info on their car.

I went with a tighter crop so that color perception wouldn't be influenced by the manila file color on people's monitors. I also color sampled the white point at least three different places on the top of each file to get a "truer" white on my color corrected monitor. Files were also converted to Adobe RGB for better color gamut. In the curves layer of Photoshop pat and I were both getting 244 R, 244 G, 244 B in the white point info display. Depending on where the sample was taken, the white could go magenta/green, cyan or red. This because the files were scanned as jpgs and the color space was sRGB (not a good way to begin).

Don't know else to tell you- my upload count came out to 376 (one intentional double entry as mentioned above).


If anyone else is looking to scan materials please use TIFF or the highest quality jpg setting possile and Adobe RGB as your color space if your scanner gives you the options. Always easier to downsample a file and then convert to jpg, do a "save as", without losing the quality of original scan.

Thanks, Howard, but the bulk of the work was done by patgreen and Stephen. I'm just detailing it out, but have to talk with Kev about exactly where all this is going, so that I can be sure the titles are formatted correctly and consistently.

Having worked on projects like the Service Index, I know that progress comes in fits and starts, often with two steps forward and one step back, and it's sometimes hard to pick up where you left off. None of us are perfect and errors are inevitible.

Stephen -

I'm not intentionaly criticizing you or patgreen in any way, but I know from past experience that projects like this can be hard to co-ordinate and the on-again/off-again nature can lead to some issues. (Heck, I was just in here, the other day, editing 48-50 Serivce Index entries, again.)

I know that the model application can be confusing to work with as Packard handled it for the 48-54 edition. I've already found duplicate etnries for same card with different model applications and will be on the lookout for more and merge the info as needed.

The problem with a handful truncated titles was not as bad as I feared. I suspect that was due to insufficient length of the image Title field, which could not have been easily anticipated. It appears that all images in the Photo Gallery get a number applied as their filenam when they are uploaded, and your filenames were then (purposely) put in the respective Title fields. Now, that I've found them all, the fix will be fairy easy.

Typos and spelling also appear to be relatively few.

Know that the missing term "Set" was only a factor with my Search criteria - wouldn't affect the database. In some cases, the word "Code" or simply the letter "C" was used, instead - though that may have simply happened in the early stages of work; some things only becoome more clear the longer you work with them. Yet, legibility of these 50+ year old tags is a problem, too. In a some cases, the tags had NO trim set info, and a few samples had no tag (left) at all.

Yet, one thing that has me confused is the syntax of these titles. My understanding is the the underscore character was intended to parse data into related fields and the hyphen (or dash) was intended to separate elements within a common field - in lieu of spaces and troublesome things like commas, quotes, and other special characters. Problem is that the usage of those characters does not appear to be correct ot very consistent. I can fix it more quickly witht he spreadsheet, but need to know the exact intent - as only want to have to do this once.

When finished editing, I am sure that BigKev's kung-fu programming skills will be able to merge the results en masse in a mater of minutes.

I'm sure all will appreciate having uniform and color-corrected images to go along with all the text data.

Brian,

The change of syntax was also Kev's suggestion. In some of the earlier posts on what to include in file title I used underscore only. Kev suggested to use hyphen within a subject description Dark-Blue for example and the underscore between subjects 2652_Cloth_Headlining_ . Turned out to be much easier on the eye to read. So the use of the underscore and the hyphen are intentional to cover the tag descriptions and for the multiple entries to search by. I guess we could call them keywords here.

Our arrival at the methodology we used was discussed back on page four and five of this topic. At that point I had to go back and rename seventy files to what they are now.

Stephen -

Thanks for the explanation and reference to previous posts.

At present we have a fine collection of images with highly detailed titles composed of keywords. As far as the site search is concerned, it doesn't matter what character the words/elements are separated by (as long as it's not alpha, a wildcard, or any troublesome special character), but with repeated reference in this thread to a database, I was thinking beyond that. With the initial suggestion of underscore versus hyphen/dash, I was expecting the title to be parsed into distinct fields, beyond a visual cue in the title.

For example, following the sample that you provided for P/N 6481031, in post #38, the actual title (which was cut-off after 'Hairlin', but I have reconstructed) appears as follows:

5622-5662-5672A-5682-5687_Set-63-73-83-113-123-163-183_Imitation_Leather_Sidewall_Dark-Green_Hairline_6481031

...and parsing, from my perspective, would render:

5622-5662-5672A-5682-5687
Set-63-73-83-113-123-163-183
Imitation
Leather
Sidewall
Dark-Green
Hairline
6481031

I was thinking that:

5622-5662-5672A-5682-5687_Set-63-73-83-113-123-163-183_Imitation-Leather-Sidewall_Dark-Green-Hairline_6481031

...would nest the elements more like the parts book:

5622-5662-5672A-5682-5687
Set-63-73-83-113-123-163-183
Imitation-Leather-Sidewall
Dark-Green-Hairline
6481031

Please feel free to correct me if I am too far out in left field.

However, there are some inconsistencies in use of the underscore and hyphen. For example, sometimes I find colors noted as:

Light_Blue

...in some cases, and:

Light-Blue

...in others

Also, I find examples like:

2650-2651-2652_5450-5451-5452

While I understand that the underscore, in this last eaxample, is used separating two ranges of models from different years/Series (similar to what the parts book does with a semicolon and a space), it would parse as two, separate fields.

Clearly, coding the tag info into an image title was no easy task, but maybe I am thinking too far ahead.

Still, I feel that all of these titles need to uniformly follow the same agreed-upon standard.

Further to the descriptions on the sample cards, I exchanged a few PM's with patgreen and BigKev and volunteered to do a database in excel format that would show additional model and set numbers for each part number that don't show up on the cards. I've done about 110 so far and have sent a copy to BigKev for any comments that he might have before continuing.

Brian