Also nailed 2 out of 5, but managed a 4154 on my first go :D
It's interesting to think about the epochal details our minds use to produce these guesses. Technology like the cars/cameras/phones present (or missing), the fashion styles on the people, the quality/sharpness of the image overall, face masks in post-2019 photos, etc.
This would be fun as a team or on a stream. Theres a lot to learn from the "speak aloud" aspect of the reasoning here. Film grain. Colour. Hair. Clothes. Cultural clues. Etcetc
The first photo I got was some guy in a cartoonish space suit talking with a kid at a festival. I guessed 1966 and got it on the nose. That heartened me.
It's tricky. I got one that was full bright standard colour... from 1912. Normal contrast a and hue and saturation, no noise. I assume recoloured later?
There are bw photos from 70s and colour photos from 50s. Contrast and noise and hue can give clues, but nothing stops a photographer from using old film in new times. I ended up going for cars and styles more when possible.
Very fun, almost wish there were less overt cultural clues though. I got 3912 on the first go but it was mostly because I recognized Bowie and Le Corbusier, and got a theater marquee with Happy Gilmore playing.
I enjoyed the pics where you needed to analyze hair/clothing styles and car models more.
I'd love to see trends and stats on the photos across regions and collaboratively filtered (e.g., someone who accurately predicts photo XX is more likely to accurately predict photo YY).
Also, it's a super curious dataset, photos with year metadata.
Simple and fun game. I just wish the score was more forgiving the older the photo was. It’s a lot harder to tell 1905 and 1910 apart than 2015 and 2020, from the perspective of 2023.
Changes in consumer-grade photo quality between 70s/80s/90s are actually quite visible if you've looked at many of them. It's subtle, but e.g. newer ones have better range in brightness, i.e. less overly bright or dark areas in high-contrast images.
Actually I've been doing consistently worse in the years I lived through than the rest of the timeline. Apparently the older I get the more styles blur together. I tend to estimate ~5 years too early for the 1980-2000, and ~10 years too early for 2010 to present.
This is a fun little game. It is interesting to see what the tells are for the time periods. The reliable ones for me so far are military uniforms, cars, and computers. Hairstyles and clothes for some time periods are a give away, but some of those have thrown me off too.
It’s incredible to see how quickly things changed ~1910-1930 in terms of infrastructure and technology. You’re going from horses and cobblestone to almost modern cars and pavement, within just 20 years or so.
Same thing with smartphones. If the photos is of a crowd, it is immediately obvious if it is after 2012 by the fact that everybody is watching their hands.
I would say that change happened much earlier, and was almost complete by 1911 in many US cities. OTOH I had a hard time telling apart cars from 1910s vs 1930s. Around the early/mid 30s a much more modern style of car came on the scene (Ford Phaeton etc. I guess), but I suppose it took a while before the Model-T style vehicles were all replaced.
Small suggestion: it would be great to get some information about the picture (location, context, people) in addition to the year, after we tried to guess the date.
Potentially, it might help find errors in the existing year-labeling - either via recognizing cases where guessers have wide divergences with a certain bias, or the explicit avenue via the about (email with errors).
Or, help identify features of a photo that definitively place it at certain times, by human perception. That might help tune autogenerated content intended to reflect certain eras to avoid anachronisms, or more strongly hint specific years or vague ranges.
Now, I don't think these are likely motivations for the game – the mere fun of playing seems enough to me. But it seems they could be.
Pretty far fetched. The date of the photos are known so ai's only intention would be to find out if humans suck at guessing or not. Which would be a useless data to improve the systems.
But some of the dates have errors, which one might, in some applications like to find.
And alternatively, how well a scene matches the audience's clear indicators of era is valuable in entertainment – even moreso than accuracy, in some cases.
Again, I don't think these are likely motivations for this particular system. But in a wide world of many aims, it's within the realm of possibiity that someone might try a system like this for such purposes.
for the first 30 seconds with the "Submit" button I thought it was a crowdsourced analytical thing though. Almost bounced. Might be worth changing the wording but otherwise agree
Reminds me of the Timeline (I think this is the name) card game where every player has a bunch of cards, each of which represents a historical event (e.g., moon landing, or sinking of Titanic, discovery of Radium, etc. Each player has to put down the card in the proper order, with respect to previous cards played. It gets harder as more cards are "on the timeline". Fun game.
Yeah, it's called Timeline. At work we took Armenian colleagues to a pub which had it available. It was funny to play with the Polish localization, they understood more than we assumed (Armenians know Russian). But the replay value is dubious. The guy who won turned out to have this game at home. Kinda cheating ;)
Except when you don't do it. My first (and best) guess was one year off, because I guessed based on a specific war with a specific way of building trenches.
It's a shame, but its privacy policy[1] is pretty clearly contravening the GDPR. There's a link on the top of the privacy page which seems to be supposed to show you a consent modal, but it doesn't work on my Chromebook, and their claimed justifications for using the user's data are clearly not actually justified.
This is really fun. My only wish is that I could see some information about the photos in the summary screen. Maybe a photographer credit or link to a Wikipedia?
I feel like this would work even better as a daily game like Wordle with the same 5 pics for everyone to play so you can (lightly) compete with friends.
And at the end, you could get some trivia for each of the photos
People like to compete: seeing the same questions makes the scores comparable. Having common context of the same "daily pics" also creates opportunities for discussion & debate which don't exist otherwise.
Trickling out synchronized games may create more durable entertainment, because otherwise many tend to play intensely until they bore of it, considering it "done". Checking back each day may generate more 'fun' per game, as well as having a built-in audience for incremental improvements/experiments that take time to implement & evaluate.
The trivia thing was asked for at reddit too, but i looked into the file storage of the server and it seems that the photos are just plainly stolen from all over the internet with file names like chronophoto.app/years/1963/ap630629032-02fcaf4ca7736109b998d760283a3a9ba7bceda7-s900-c85.webp
or
chronophoto.app/years/2017/ed-jones-north-korea-top-100-photos-2017.jpg
No credit is given anywhere too. This is probably illegal and the author will get in a lot of trouble if he does not do something about that soon.
Neat simple little trivia thing, but fun to think about each one and any clues. I got 1969, 1968, 1971, 1906 and 1904, score of 3435. First two were war related, for 1000 and 777. First was b&w but obvious forest clearing mid Vietnam war, so bit of a guess but on the dot, then the second was a color picture of a bus with "Saigon" on it that I underestimated but again, era was clearly Vietnam. Second one showed people in front of cricket scoreboard and with all the batters names and that was pure luck for me because only one possible famous match featured those players! (test #679, Australia v Sydney 1971). Utterly bombed the next, 83 points, I agree with others that I found older photos tougher. Guess 1923, way off. Then I could guess at the final based on the period late 1800s/early 1900s dress at 1910 which was good for 575.
Anyway, fun in terms of trying different strategies and it making one focus on the details whereas normally with photographs it's easy to just sort of look at it more abstractly without really thinking about it. Could be a neat party game even as-is, but I also agree with a sibling poster that it'd be easy to expand a touch to favor that (allow having n different guesses each with different colors, so everyone can take a shot for each image for some light competition maybe?).
So fun! Have some friends courious about culture and history that will love this.
I think the UX can improve a bit:
- Would like that after a submit show both the correct date and the player guess, now only shows the correct date on screen (and the guess in the slider but not with numbers, just a visual gap)
I got a 1976 and a 1992 correct on my only attempt. What are the odds of that? You can narrow a picture down to roughly a 20 yr span. Is that 1 in 400 for just those two? Worst case would be 100 x 100?
2,889 on my first round, 4,4xx on my second. It's fun zooming in and looking for cars, logos, and clothing style for hints. It's also neat to see, say, a road with cars from Model T style through to swoopy 1930s sedans. Contrast that with most historical movies where, if it's set in 1955, every car is from 1950 - 1957.
A good game for kids to show them how to recognize different eras. Many clues in each pic, including the colors that are representative of the film/tech used.
As my last image in round 5 I got a photo of WWF where Mankind was on top of the hell in a cell. I guessed 1998 because of the meme and got it correct! It was the only image I hit dead on.
Same for me with Musk on stage with the Cybertruck (2019).
It would be more interesting if these popular photos were removed, and guessing would be based only on knowledge about clothing and hair styles, film type, car design, etc.
Absolutely. It would help if the results included some context for the photo, but that seems like a great way to spend educational & fun time with your son.
Just beware there seems to be little content filtering, one of the images I got was pictures of war casualties tossed on the side of the road from 2015.
This is very fun, but I'm wondering why you haven't included source or some context for the photos? Also, you've disabled right click so reverse image searching them is painful.
My issue is there being no credits or sources for the image. Screenshot and reverse image search is certainly a solution on my end to get to that info, but it should be present on the website, probably at the round summary.
I recently been going through all of my family photos. A lot of them are scanned and undated.
I’ve been looking for some kind of machine learning library that will allow me to automatically date them. Approximates are fine, but an algorithm would do better than I would.
Even just grouping photos of similar film stock and scene would be useful as they’re all completely jumbled.
Does anyone know if there’s anything like that?
Trying to use this on mobile sure is frustrating. I pinch zoom to look at the photo and somehow I end up changing photos when I zoom back out to click the buttons.
My scores are getting progressively better and I’ve reached 4481 by the 5th game!
I guessed better with photos that had telling elements such as hair styles, clothing, cars, beards - and preferably more than one instance of them. Bigger samples got to better results.
Fun game overall. OP, thanks for sharing the link.
I came back to play again after 10 hours and noticed some of the 1950's photos are fake colored and fooled me. I think those photos make this game less authentic.
I feel I'd be doing better at this game if I was a car person. The only time I've got the year spot on was with a photo of the toppling of a statue of Saddam Hussein. The worst I did was guess 1960 on a black and white photo taken almost directly above of some cars on a road, it was actually 1980.
First two games I tried I did pretty well on (~3000 score?), then had some shockers (at least 2 or 3 I'd need some convincing on, photos dated in the 30s that were in full high resolution colour?), then got lucky with 3 photos clearly in the early 1900s, one with a cinema listing a film I knew the release year for and one with a somewhat iconic pre-war coca cola ad that I managed to pick the exact year for, scored just over 4400 (I guess 5k's the max?).
Willing to bet a suitably trained AI would whoop any human's ass though, even ensuring none of the photos were explicitly part of the training data.
Perhaps, but this particular one (looked like some dignitary inspecting an army, assuming an Asian country but couldn't say which) was especially crisp with high colour saturation that didn't match that of other photos from that period, indeed I assumed it was from the 70s or 80s.
https://jsfiddle.net/9dvx1sua/show is another photo I would never have guessed the year of from the quality of the image. Even if the outfits/equipment are obviously from a particular period, I'd think it was a reenactment or a scene from a movie...
It would be nice to have a link with more information about the single photos, at least the exact time and place where it was shot, i can then look for the related events myself.
I got 2827 as a score which isn't that great, but I nailed the year perfectly on 2/5. Super interesting!
reply