What is the algorithm used by Googles reverse image search (i.e. search by image)? What algorithms would I need to understand, to create similar functionality on a small scale?

Google image search

It depends on what features of Google Image Search you are thinking about: the retrieval or the classification/recognition.

Anyway, you can just get some general ideas about the main technologies that they are using.

nWhatever the method or descriptors, rest assured that there is a lot of Google flavored optimizations and tweaking under the hood.

I will be referring here only to the visual description.

Clearly they are fusing multiple types of descriptors and information in the final product.

,ClassificationnFor the classification/recognition problem where you want the system to tell you what objects/landmarks/people it can recognize in the image that you feed to it, they use, most probably, Deep Learning as mentioned in the other posts.

The deep learning experts that have joined them (Andrew Ng and Geoffrey Hinton & Co.

) surely had some words to say on this (although they must be working also on other non-vision problems in the same time).

Once the image is classified (known objects or concepts were detected in the image), the image labeled and then introduced in a text search engine framework making it possible to retrieve the image by tapping some keywords.

,RetrievalnConcerning the retrieval part where you perform search/query by image, I believe that the technology that they use is more or less the same one as in Google Goggles.

Probably they use a variant of the Bag-of-Words (Page on Ox) or VLAD (Page on Hal) with their own interest point descriptors (adapted local Gabor wavelets), clustering, quantization and dimensionality reduction methods.

There is a talk given by Hartmut Neven (head of Google Image Search team) at ICML in 2011 that might be worth watching Machine Learning in Google Goggles The first part of the talk gives some info about Google Goggles, while the second part about Quantum Optimization might be relevant in 10 years from now :) ,nWith the big progress made by deep learning recently, making it possible to crunch a lot of data and learn from it, I think that the retrieval and classification tasks (until now tackled as different problems) should get closer as they are quite complementary.

Especially when you must handle so much data.

I see them used together for joint image retrieval, discovery and labeling in a more automatic manner.

Search by image

I will answer with a lot of screen captures (in French, sorry!).

,Go to Google Images:,Clic on the share button on the bottom:,Choose the Desktop view:,The Google page should be like this:,Clic on the camera icon, you will have this:,Now you can import an image from your photos gallery (or maybe from one of your cloud):,Thatu2019s all!

Face search person by photo

Google Images (Google Images) lets one upload a photo and search for other photos of the same person.

,As an experiment, I uploaded this image of myselfu2026,u2026and Google both identified me by name and found this other photou2026,Interestingly, it also found this gentlemanu2026,u2026coincidentally also a lawyer, apparently because his photo happens to appear on my FounderSpace page and Google thinks he and I are the same person!

How to find the name of a person in a photo

Use reverse image search.

Following are the best tools for the purpose:,TinEye,Google,Baidu,Now a days reverse image searches have become extremely powerful.

Once you do reverse image searches, you will find interesting information.

How to find someone with a picture on Google

Reverse Image Search.

Google is good, but I like to use u201cTin Eyeu201d search engine.

,Do you have a printed photo, or digital image file?,Reverse search will only work if the image has ever been on World Wide Web.

Google image search face

North Koreans are originally from old Goguryeo and Balhae Kingdom while South Koreans are from Silla and Paekche, while Seoul together with Gyeonggi-do is a mixture of them all.

,North Koreans are more slim in face due to malnutrition (not Kim Jong-Un of course).

,North Koreans donu2019t smile until they are told to do.

,North Koreans have a more uniform hair style.