Identifying PO Wagon colour

[wagonman]

On 02/05/2023 at 08:48, Compound2632 said:

PO wagons fall into the following categories:

1. those for which there is documentary evidence and a photograph including a record of livery (e.g. Gloucester officials);
2. those for which there is documentary evidence and a photograph;
3. those for which there is documentary evidence (e.g. railway company PO registers, builder's records);
4. those for which there is only a photograph;
5. those which are known to have existed but for which there is neither documentary evidence nor a photograph;
6. and those which existed but have left no trace.

 

The last two categories seem to be the commonest. Class 5 can be deduced from the owner's membership of the RCH commuted charge schemes if after 1926. Sometimes a wagon will be mentioned in an accident report or a repairer's notebook – but I suppose that counts as documentary evidence.

[kevinlms]

7 hours ago, Bristol_Rich said:

AI comes up with these results…as an indicative guide…

 

Finally, a proper usage for AI software!

 

Perhaps it's biggest error is to assume that the brake lever and some parts of the brake gear, would match the body colour.

[Bristol_Rich]

5 hours ago, kevinlms said:

Finally, a proper usage for AI software!

 

Perhaps its biggest error is to assume that the brake lever and some parts of the brake gear, would match the body colour.

In the absence of any records or original paint diagrams, it’s a useful tool to provide an indicative coloring.

 

if you pushed the image through three different AI tools it’ll highlight the consistency and inconsistencies between the tools and provide a fairly reasonable representation.

[Compound2632]

2 hours ago, Bristol_Rich said:

In the absence of any records or original paint diagrams, it’s a useful tool to provide an indicative coloring.

 

if you pushed the image through three different AI tools it’ll highlight the consistency and inconsistencies between the tools and provide a fairly reasonable representation.

 

I'd want to try it out on a large number of monochrome photos of wagons where the colours are known before relying on it to any extent. And then it has to be intelligent enough to understand the effect of factors such as type of emulsion (orthochromatic / panchromatic etc.) and level of gloss of the paint surface. 

[wagonman]

The only thing I'd commit myself to is that the lettering is shaded. If the diamond on the door is indeed red then it's unlikely the rest of the body would be the same colour which points towards grey (or dare I say green? No, I daren't) as the most likely colour. Perhaps you could ask your AI robot to try that – and rusty black for the underpinnings.

 

 

[Bristol_Rich]

 

6 hours ago, wagonman said:

The only thing I'd commit myself to is that the lettering is shaded. If the diamond on the door is indeed red then it's unlikely the rest of the body would be the same colour which points towards grey (or dare I say green? No, I daren't) as the most likely colour. Perhaps you could ask your AI robot to try that – and rusty black for the underpinnings.

 

 

I’ve intentionally noted twice it’s indicative and a ‘guide’ and not wholly accurate…however in situations where there is an absence of any detailed guide or historical record then unless you want to take a wild punt, its there for your own personal use so fill your boots… I don’t need to ask it to anything thanks👍

[Compound2632]

7 minutes ago, Bristol_Rich said:

I’ve intentionally noted twice it’s indicative and a ‘guide’ and not wholly accurate…however in situations where there is an absence of any detailed guide or historical record then unless you want to take a wild punt, its there for your own personal use so fill your boots… I don’t need to ask it to anything thanks👍

 

Understood. I just think it would be very interesting to try it out on examples where the answer is known.

[JohnR]

9 hours ago, wagonman said:

The only thing I'd commit myself to is that the lettering is shaded. If the diamond on the door is indeed red then it's unlikely the rest of the body would be the same colour which points towards grey (or dare I say green? No, I daren't) as the most likely colour. Perhaps you could ask your AI robot to try that – and rusty black for the underpinnings.

 

 

 

Quite, its very interesting that AI has suggested it is red - I've been told that Red Oxide (or variants thereof) were not uncommon for co-op traders wagons in Scotland. But given that the central diamond is a dark red colour, according to Dundonians who recall the co-op in later years (from the cover of the divvy book), I'd say it was unlikely the body colour would be red as well, even a lighter shade.

 

Now if these AI tools allowed you to choose a colour for particular areas, and then filled in the rest in comparison, they'd be VERY interesting. Or if you could, as @Compound2632has suggested, even train them on B&W images where the wagon body colour was known....

[wagonman]

I/we am/are not trying to diss AI but to improve it. I would be delighted if it could be made to work with a reasonable degree of accuracy, but can't think how this could be done. I heard recently of a case of AI being used to distinguish between skin moles that were or were not cancerous. Unfortunately it "learned" that anything with a ruler in the photo was cancerous...

[Wickham Green too]

1 hour ago, wagonman said:

... anything with a ruler in the photo was cancerous...

Better not aim it at any pictures of todays Coronation, then ! 😊

[roythebus1]

when I worked at the York Way model shop in 1968, the shop manager AG Thomas produced a series of books, "A modeller's sketchbook of private owner wagons". these were, as titled, sketches of PO wagons where he showed the style of lettering and colours. I found one of the books during a sort-out yesterday. If I can find it after the sort-out, I'll have a look in there and see if this wagon is noted.

[Compound2632]

15 minutes ago, roythebus1 said:

when I worked at the York Way model shop in 1968, the shop manager AG Thomas produced a series of books, "A modeller's sketchbook of private owner wagons". these were, as titled, sketches of PO wagons where he showed the style of lettering and colours. I found one of the books during a sort-out yesterday. If I can find it after the sort-out, I'll have a look in there and see if this wagon is noted.

 

A.G. Thomas' three Modellers' Sketchbooks are included in Joe Greaves' index:

https://lightmoor.co.uk/BDLpdf_files/Private_Owner_Wagons_Index.pdf

I'm afraid this wagon / firm is not listed.

 

Those books do seem to be hard to come by second-hand.

Edited May 7, 2023 by Compound2632

[roythebus1]

Ok, thanks, that'll save me a bit of time trying to find something that isn't there. My AG Thomas books are signed by the author as well. :)

 

[Wickham Green too]

Now THAT's something you won't get in the internet - a Signed First Edition website !!! 😊

 

( I do have signed copies of Bill Hudson and Keith Turton volumes - but they both signed freely at the Warley show etc.)

[Caley Jim]

No nearer providing an answer, but i have put a query on the Caledonian Railway Association forum and will report back.

 

Excuse me if I retain a degree of scepticism over AI colourised photos!

 

Jim

[Wickham Green too]

A is for Artificial .....

[JohnR]

6 minutes ago, Caley Jim said:

No nearer providing an answer, but i have put a query on the Caledonian Railway Association forum and will report back.

 

Excuse me if I retain a degree of scepticism over AI colourised photos!

 

Jim

 

I think AI is ok at things like colouring grass green, faces pinkish, etc - predictable colours. 

[Wickham Green too]

3 minutes ago, JohnR said:

I think AI is ok at things like colouring .. faces pinkish, etc ...

Can it differentiate between white European and Afro Caribbean faces ...... or little green men ?

[57xx]

On 05/05/2023 at 00:34, kevinlms said:

Finally, a proper usage for AI software!

 

Perhaps it's biggest error is to assume that the brake lever and some parts of the brake gear, would match the body colour.

 

That and the fact the repaired top plank would most likely be unpainted. It looks about as accurate as a colourised postcard.

[Wickham Green too]

3 hours ago, 57xx said:

... about as accurate as a colourised postcard.

Generated using AI .... = Allegedly Intelligent. 

[JohnR]

19 hours ago, 57xx said:

 

That and the fact the repaired top plank would most likely be unpainted. It looks about as accurate as a colourised postcard.

 

Top plank cant be unpainted, as it has the original lettering. 

[Compound2632]

10 minutes ago, JohnR said:

Top plank cant be unpainted, as it has the original lettering. 

 

That's missing the point that the top half of the top plank has been replaced and it's that replacement part that is probably unpainted. At least, that's my interpretation of the photo. I don't think the original suggestion that there's material folded over the top of the plank is correct.

[57xx]

4 hours ago, JohnR said:

Top plank cant be unpainted, as it has the original lettering. 

 

Look closer, it only has half the lettering - the bottom half. As Stephen says, I'm referring to the fact the top half of the top plank looks to have been replaced.

[57xx]

Here's one for @Compound2632, MR's well known maroon livery.

[Dave John]

The ai results are interesting, I have had a play with some of the sites myself. As JohnR says , the ai is good at identifying and colouring grass and faces. 

 

The ai is learning from the internet, there are vast numbers of colour pics which include grass and faces it can learn from. How many colour pictures of railway wagons prior to the 1960s are there? Very few, though I suppose correctly painted preservation stock might be included. It does seem to recognise a railway wagon and that they tended to be browns and greys. 

 

We would recognise a MR wagon and tend to believe that it was grey, or a Caley wagon and know that it was red. However we have spent years reading about and discussing these things across a range of media that the ai doesn't have access to. Nor can it relate the letters MR to the Midland Railway. It certainly can't relate the Dundee co-op to any specific colour, but then neither can we.

 

Ai is still in its infancy, it is still learning. Give it a few years and it might be able to identify a MR wagon and know that it should be grey.

 

Mind you, ask it to correctly colour an early 20 C GWR one and it might just blow a fuse.