I am interested in how google have done it's image search using an image. I am doing a project on image comparison and searching using images, and considering how impressive google image search using image seems to be, I am curious what they have done.
I also wonder how much they do actually image comparing and how much they use like keywords around images etc.
Related
So, I don't believe that exists someway to "read" a PNG through its binary code, or something like that, but I have zero knowledge on Image Processing or Computer Vision, and so, I can't be sure if that are ways to do it or not.
To be clear: I want to know if there are ways to identify the image of a Logo using an image of the Logo as reference but through methods that use only the binary of the image.
Thanks in advance
If the logo is known, and not distorted a lot (logo printed on a scarf is not as good as on a flat surface), there is technologies that can achieve that:
It is a template matching problem, see https://docs.opencv.org/4.5.2/d4/dc6/tutorial_py_template_matching.html.
I'm trying to automatically process the results of a google similar image search. I've noticed that some results come back with "x-raw-image:///<48 bytes or so of data>" in the URL for the image. I think this is happening when the search is extracting an image from a PDF document. Does anyone know what that data is? Is it anything that is usable outside of google? I was hoping that it might be similar to a "data:image" reference, with the image data embedded, but that might not be the case.
Thank you.
ETA: Here is a random example I found that I hope works for everyone:
https://www.google.com/imgres?imgurl=x-raw-image%3A%2F%2F%2F79da01aca79b556defc81ee719442327379e067a26c5c5c9ec104fc39fb70177&imgrefurl=https%3A%2F%2Farxiv.org%2Fpdf%2F0706.2165&tbnid=bVWbp6wB2nmBiM&vet=12ahUKEwiD06O97dXuAhWUUM0KHXp7Co0QMygCegQIARAy..i&docid=nD--BiC0InIh_M&w=1600&h=2071&q=physics%20papers%20diagram%20pdf%20x-raw-image&client=firefox-b-1-d&ved=2ahUKEwiD06O97dXuAhWUUM0KHXp7Co0QMygCegQIARAy
This is the frame you get when you click on one of the results in an image search.
I use Google Cloud Storage for images. Images get uploaded to my Google App Engine based server. To serve the images I use the feature that allows image scaling. Just appending =sXXX to the image scales it to the desired XXX size for the largest side. Pecurliar that omitting =sXXX suffix scales to 512 pixels.
In order to NOT scale, one has to add =s0. This is what I do.
Sometimes, not always, I get this interesting problem. Serving URL uses =s0 suffix to ensure the original image gets download. Here is the example, which works
bread image, original size, =s0 suffix
It should work, this is documented in this Google App Engine article:
Images Python API Overview and also in this article Uploading, Resizing and Serving images with Google Cloud Platform
However for some images it does not work. Example:
This does not work:
http://lh3.googleusercontent.com/Kwj_Fe938vOxsVw9KFILW-MaxoNJlGinxfAHqj4_3bSxhHWPoSLXb5zXGn5k7V4BJgWRlgD1zvULpmjCnFhE1Q=s0
However removing the suffix =s0, shows image all right.
http://lh3.googleusercontent.com/Kwj_Fe938vOxsVw9KFILW-MaxoNJlGinxfAHqj4_3bSxhHWPoSLXb5zXGn5k7V4BJgWRlgD1zvULpmjCnFhE1Q
At least some other sXXX suffixes work
http://lh3.googleusercontent.com/Kwj_Fe938vOxsVw9KFILW-MaxoNJlGinxfAHqj4_3bSxhHWPoSLXb5zXGn5k7V4BJgWRlgD1zvULpmjCnFhE1Q=s250
So far I have not found a pattern. Looking at Google's bug repository for Cloud Storage does not reveal anyting useful.
Anyone has an idea where to look?
Google PageSpeed Insights is saying that I should compress and resize several images.
Here's an example:
I saved this image using Photoshop's Save for Web JPEG High preset:
What can I do to this image to avoid Google's "compress and resize" warning?
Due to the small number of different colours that image seems to have, you'd probably get a much smaller image if you exported it as a gif, or png-8. Sitepoint has a useful article on when to choose which image format.
This is what works for me and JPEG to be on the safe side of PageSpeed Insights.
File > Save for Web & Devices
Not only do you need to find the right compression settings but also you need to stick to your img tag actual dimension. If you know your max width for img tag will be 555 px then save your JPEG to that dimension, not a pixel more ... as for how it goes along with retina display, it seems Google does not take this criteria in account right now.
If you are not happy with the resulting quality (and you might be if you have a designer ecommerce site) you need to bypass Google recommendation - knowing it could slightly penalize your SEO over someone else that actually stick to the rule.
You don't mention if you are using the Chrome extension or the main website? Assuming the former you can actually download the optimised version of an image by clicking the 'See optimised content' link under 'Suggestions for this page'.
Img here:
I've looked around, this question has been asked before but i would like to know if it was possible to do so, maybe without using an API, or using another search-engine.
I would like to check the web for images, by searching with an image.
Is there any way to do so, if so can anybody point me in the right direction, i just can't seem to get started.
You can do this currently with Google. Drag and drop the image into the search field.
Just go to:
http://images.google.com
and drop your image over the search field.
Google used to have some images below the search field that you could drag and drop, as examples.