- Видео 4
- Просмотров 46 956
Yusuke Matsui
Добавлен 9 авг 2010
【東大オープンキャンパス2020】高校数学で理解する画像検索
Google画像検索のようなサービスでは、画像をアップロードすると、それに似ている画像をインターネット上から探すことが出来ます。このように、似ている画像を探す技術を「画像検索」と言います。本講義では、画像検索において速度・精度・メモリ消費量がどのように関連しあっているかを高校で扱う範囲の数学のみを用いて解説します。これにより、実世界のシステムを記述するためには高校数学の知識が重要な役割を果たすことを体験してもらいます。本発表は「高校生のための東京大学オープンキャンパス2020」において、「工学部 電気電子工学科・電子情報工学科」が分担した、公開講義です。
・高校生のための東京大学オープンキャンパス2020:cdn.pr.u-tokyo.ac.jp/
・スライド:bit.ly/3hUxgW8
・東京大学工学部電気系(電気電子工学科+電子情報工学科):www.ee.t.u-tokyo.ac.jp/j/
・松井勇佑:yusukematsui.me/index_jp.html
スライド中の漫画画像のクレジット:ARMS, (c) Kato Masaki, Manga109
・高校生のための東京大学オープンキャンパス2020:cdn.pr.u-tokyo.ac.jp/
・スライド:bit.ly/3hUxgW8
・東京大学工学部電気系(電気電子工学科+電子情報工学科):www.ee.t.u-tokyo.ac.jp/j/
・松井勇佑:yusukematsui.me/index_jp.html
スライド中の漫画画像のクレジット:ARMS, (c) Kato Masaki, Manga109
Просмотров: 616
Видео
[CVPR20 Tutorial] Live-coding Demo to Implement an Image Search Engine from Scratch
Просмотров 31 тыс.4 года назад
[CVPR20 Tutotrial] Image Retrieval in the Wild matsui528.github.io/cvpr2020_tutorial_retrieval/ Live-coding Demo to Implement an Image Search Engine from Scratch Yusuke Matsui slide: speakerdeck.com/matsui_528/cvpr20-tutorial-live-coding-demo-to-implement-an-image-search-engine-from-scratch code: github.com/matsui528/sis demo: www.simple-image-search.xyz/
[CVPR20 Tutorial] Billion-scale Approximate Nearest Neighbor Search
Просмотров 14 тыс.4 года назад
[CVPR20 Tutotrial] Image Retrieval in the Wild matsui528.github.io/cvpr2020_tutorial_retrieval/ Billion-scale Approximate Nearest Neighbor Search Yusuke Matsui slide: speakerdeck.com/matsui_528/cvpr20-tutorial-billion-scale-approximate-nearest-neighbor-search
DrawFromDrawings: 2D Drawing Assistance via Stroke Interpolation with a Sketch Database (TVCG 2017)
Просмотров 7387 лет назад
Yusuke Matsui, Takaaki Shiratori, and Kiyoharu Aizawa, "DrawFromDrawings: 2D Drawing Assistance via Stroke Interpolation with a Sketch Database", IEEE Transactions on Visualization and Computer Graphics 2017 (in press). Project page: www.hal.t.u-tokyo.ac.jp/~matsui/project/drawfromdrawings/drawfromdrawings.html
Awesome presentation, thanks!
Very well explained! Thank you!
when i run code it says internal server error. how can i fix it
thank you for your demo
Hi. What did you click to access the link at 19:44? Please can you reply ASAP?
Sir, can i use this code on android studio apps? Its kinda mobile application
What algorithm is used? to process the searching
Hi I would like to build something similar with you demo. Is there any chance to discuss about a small project? Lmk!
Cool stuff.
Cool stuff
But is it possible to scale this up for lets say 1k images , wont it slow down?
very helpful video sir, but when i upload different data from datasets it showing results but i want is None as output
then i think you should specify the criteria of output probability. if it is less than 30% then eliminate that O/p
Awesome, You are the superb coder I ever seen, Thank you soo Very much
I was curious and googled your name, then found that in result page, wiki had introduced you as a baseball player 😆
Thanks Mr Matsui, it is a great video!
Is there a way to upgrade it that the results will show links of websites they are extracted from
How to run
If I want to apply and check the performance of different models then How can I do it?
Fast api
How to use Django instead of Flask?
At 11:44, dist = q_norms[m] + x_norms[n] - **2** x ip[m][n]? Shouldn't ip[m][n] be multiplied by 2 (to match the formula on the top of the slide)?
Thank you for your explanation!
Thank you! That is awesome.
I expect that we can do better than kmeans clustering for dimension reduction and coarse quantization, for image data at least, by smart feature detection using transfer learning like ResNet50 as main body, plus VAE as head of network for smart dimension reduction. Also use built-in tensor quantization on the short vector, or build your own quantizer that is differentiable for purpose of backprop by custom sequential Relu activations that start at zero, which is stabdard relu, then 1,2,3,... Until it swept entire 8 bit range ie 256, or 16 bit range ie 16k. No more fine search is needed consequently, just direct addressing to correct hash bucket, and pull out an item from its list. Or you can search fine list other ways like pq or lsh .
This presentation is great. I love it! It's a fast way to understand the summary of the state of the art of large scale search.
Can you create a data base like this for me? I would pay
Really very nice project and easy to run and learn and hasslefree
Thank you so much.
Brilliant Project !!
Thank you !!! very clear
Thank you for the demo, very helpful!
I have an app built in react native using javascript. Is there a library like this that I can use in my app? I would like to have it looks up local storage for images not online.
The submit query button is showing all my images data....not the related images as query Can anyone help....
On simple-image-search.xyz do you have millions of image in your static/img database ??
Heyyy, please help me. How can I make this without having to own 100 images. can't I use Imagenet's API or something to show results online ? I want to deploy this on heroku but don't want to put all the millions of images in my img folder
Hey, did you deploy your project in heroku? I also need to deploy in heroku.
Daiiiiiisukiii
Thank you for such a great video I really enjoyed it
I love you
Very well explained topic. And a great presentation as well with nice colors for the hashing function and visual on the coarse graph to fine graph slide. Thanks so much
This is fantastic. Thank you so much for posting this. This is easy to follow and it works exactly as expected. Great job!
Hey bro...... I m getting the images from my data(100 images) as the query output...but not the related from search images Can u please help.
Thank you for a great summary!
ありがろうまついせいんせい
An excellent work.