Doing a search for an image on google images

Can someone explain to me how this works?

I have tried following the directions, and i haven’t had any luck. Perhaps the photo I was using was in a location that wasn’t accessible by google, or perhaps I just don’t know how to search on an image properly.

Also, when you do an image search, does the image have to exactly match? And if so, how does google’s software match a photo? If not, will a google image search return photos of the image that is in the photo I submit, without it being the same photo?

For example, if I do an image search on a city that I don’t recognize but I would like to know what city I am looking at, will the search, if it works properly, return the exact same photo, or will it return other photos of the same city?

There are two ways I’d like to know how to search.

  1. the photo is on my hard drive.

  2. the photo is on a web browser, but not downloaded to my hard drive.

Thanks for any help!

Start here:

How to search by image

For uploads or drag and drop, Search by Image - includes a video tutorial

It sounds like what you’re looking for is the reverse image search, not the regular image search. When Google does it, it’s mostly based on the text around the image on webpages, so it won’t work so well for an image from your hard drive. There’s another image search site called TinEye, though, that looks at the image itself, and somehow (their algorithms are secret) finds other images that “look like” it. That’ll work fine for something from your HD.

No, both companies do image searches (meaning they look at the picture and find similar ones, regardless of nearby text – otherwise you wouldn’t be able to search for a picture you upload yourself). There are a variety of algorithms that do this, but basically it breaks the picture down into sectors, reduce the information in each sector (coming up with a sort of hash or fingerprint for that sector), and then compare the sectors across its database to find a % match for other images that share a large number of sectors. It can find exact matches or even similar-looking images that share roughly the same shapes and color palettes (like try searching for a picture of a face or a sky).

It used to be that Google couldn’t do what TinEye did, but that hasn’t been the case for quite a while, and I think Google’s database is quite a big bigger now.

There are a number of Firefox extensions, any one of which will add a Google Image Search item to the right-click menu, and open in a new tab.

Google, as with most of it’s products, keeps changing Google Images for no improvement.

Never mind that. Google doesn’t know what your image looks like and doesn’t look at it. It only searches for what you tell it. Use keywords. Describe what you are looking for using several words. Try some related words.

As for “can it identify the city from a photo”: it depends.

Google has no intrinsic awareness of “this building is from this city”. BUT it does have a huge database correlating pictures with text on webpages. If 30 different travel websites use the same photo of the Empire State Building, Google knows that that picture is associated with the building. When you upload that same photo, Google can tell you about this.

But if you upload another shot of the Empire State Building, taken from a different location or angle, it is extremely unlikely that Google will know what that building is – unless it’s visually so similar to another preexisting shot of the building that its algorithm conflates the two.

In other words, Google knows about pixels in images and the texts on webpages that surround that image. It doesn’t yet have the ability that humans have to be able to discern an object in imaginary 3D space, spinning it around until it fits some pre-existing concept of the Empire State Building. For example, if you rotate a face upside down or flip it left and right or crop it just right, GIS probably won’t find it.


On the other hand, Microsoft Research is working on technology that clusters photos together to make 3D models of a place by clustering similar photos and overlapping them on top of each other and calculating the perspective shifts in each one. So if you feed the software 3000 tourist pictures of the Notre Dame cathedral, it’ll generate a photorealistic 3D model of that place from all the pictures.

This doesn’t quite extend to image searching yet, but once the information is there, it ought to be possible to match your uploaded photo to something already in their geometry-adjusted databases. I wouldn’t be surprised if that was their next project for Bing… so keep your fingers crossed.

No, that just isn’t right. Go to images.google.com. In the search box, click the camera icon. You can upload your own picture to search for or feed it a URL of an existing online picture. It will then search for other occurrences of that picture without any additional search terms.

Google Image Search does BOTH. It can search by keywords OR visual similarities.

Yes it does. However, unless it finds the exact same image, the matches it comes up with are very poor.

OK, then Stink fish Pot is going to have to upload the comparison photo or link to it. Either way, we’re a little ways off from computers that understand what the fuck we are looking at. Keyword search might be as good or better.

Exactly. And unless we know how it defines the “exact same image,” we are unsure of the results.

Google is great at finding other occurrences of the same picture, if you’re trying to find out what website the picture came from. But if you’re searching for a picture that you took yourself, then it isn’t likely to find an exact match.

If you’re scanning old family photographs and trying to use Google Images to find out where they were taken, there’s almost no chance of getting a match. A better bet would be to examine the photos for business names or street signs or other clues, and do a keyword search on them.

Very true. The use cases are different. GIS is almost always useless for finding different pictures of the same thing unless you use a keyword search, as you pointed out. (EDIT: Unless the photo you upload is already a commonly used one. If you upload an unnamed but common picture of a celebrity, for example, Google would very easily be able to tell you who that is and show you other pictures.)

It is, however, very useful for finding out which other websites use a certain picture, where a picture came from, whether there are lager versions of it, etc.

In practice it actually works pretty well unless a picture was deliberately manipulated to not show up on GIS. Give it a small picture and it’ll find you a bigger one, give you a meme and it’ll often find you the original picture without the text, etc.

I don’t think Google has specifically discussed their particular algorithm, but in general search-by-pixels is a growing field and shouldn’t be altogether dismissed as useless.

Take a look at the Photosynth link I posted above. Already, consumer-grade software can look at multiple pictures, see where they overlap, and reconstruct a scene by stitching them together in 3D space. This isn’t quite object recognition, but it’s still very impressive in that it has to understand light, shadows, angles, perspective etc. from a bunch of colored dots. It hopefully it won’t be long before you can feed it a picture of a city and it’ll interpolate your photo between pictures it already knows about and deduce that this is likely angle X of street Y in city Z…

Anyways, generally speaking, if Google can’t get the image from the URL, then you need to download it to your computer and then drag the file into the text box.

The analysis seems to be based more on the colors in the image than their exact arrangement, so anything other than fairly close versions will not show up. However, Google will also include the keywords it thinks have something to do with that image, and you can change those to tailor the results.

If you are looking for the exact same image, and Google isn’t finding it, then it pretty much means that Google doesn’t have that location in its image index. As stated above, you might have better luck with Tineye or other image search engines.

In my experience, what Tineye does, it does significantly better than Google. Honestly, I’m a bit surprised that Google hasn’t yet bought them out.

Thanks for the information, everyone.

Exapno Mapcase, I actually read the directions you linked to, and that’s why I started this thread. I still haven’t been able to get this to work, but I am not convinced that I am doing things properly.

Can I link to a picture on a website and do an image search with that? Or is it possible that the image on the page cannot be linked to?

Instead of a city, let’s use a person. I see a picture of someone and think they look familiar, but can’t place them. If the website won’t let me copy the picture, then I guess it’s possible that it will not be able to be linked to in the search, correct? So in this instance I am out of luck. But if I can copy the picture to my hard drive, then I could put the picture in and search for it without keywords, correct?

I should mention that I don’t want to use keywords in the search. I thought the whole idea of an image search was to match a picture with others that are found on the net. Obviously, if. I want a picture of a certain actor or city, I can put that in google and do an image search. That works fine. I’m actually trying to understand how I search on a picture by itself, and have the image search return images that match the person in the photo without having to know the name of the person or object in the photo.

I have yet to have a successful search on just an image, so I’ll re-read the directions and re-watch the tutorial again and see where my fatal flaw is.

Go to images.google.com. Click the camera icon in the search bar. Either give it a image URL or upload an image from your computer.

Some websites will try to prevent you from copying an image. In that case it gets more complicated. If you have a specific website in mind, show us… and tell us which browser you’re using.

Ah!

I think I found out why I’m having so much trouble.

I am working from my iPad, and when I go to image.google.com, there is no camera icon.

I don’t know why I didn’t remember that the first time I watched the video, but I thought I just wasn’t looking for the camera icon in the right place. I think that is because the camera icon should be on the image.google.com page, and. I shouldn’t have to hunt for it at all.

I was trying to put the url into the search box, or the location of a picture on my hard drive in the search box on the google page. But that’s incorrect. I need to find that camera icon and click on it, and then use that to insert the image.

In the video, it shows a camera icon two different ways. One, it is in the search box at the very end, and takes the place of the magnifying glass icon. Two, it shows an icon of a bigger camera standing alone outside the search box. Either way, I don’t think I have the ability to run it from my iPad.

I don’t understand why, but it would seem at least that I’ve discovered my problem.

When i go to images.google.com, i get a google page with the word images under the word Google, but no camera icon. Why is that?

Am i on the right track here?

iPads get different webpages served to them. Do a “request desktop site” if you have that option in your browser. If not, try the first answer in this article:

Or just do the search on a computer.

My experience is exactly the opposite. I used to use Tineye, now I find Google reverse search does it far better.

Just FYI for the future, it’s really critical that you tell us if you’re using a mobile device or tablet. They get served an entirely different page than you see in a desktop browser, which is why Google’s instructions didn’t match. (Do iPads even have hard drives you can access from a browser?)