Content-Based Image Retrieval using the Bag-of-Words Concept |
| Английская грамматика | ||
| << Part 1: Bag-of-words models | Класс сгате >> |
Автор: mturk. Чтобы познакомиться с картинкой полного размера, нажмите на её эскиз. Чтобы можно было использовать все картинки для урока английского языка, скачайте бесплатно презентацию «Content-Based Image Retrieval using the Bag-of-Words Concept.ppt» со всеми картинками в zip-архиве размером 1878 КБ.
Content-Based Image Retrieval using the Bag-of-Words Concept
содержание презентации «Content-Based Image Retrieval using the Bag-of-Words Concept.ppt»| Сл | Текст | Сл | Текст |
| 1 | Content-Based Image Retrieval using | 10 | Extraction. Represent each patch/interest |
| the Bag-of-Words Concept. Fatih Cakir | point with SIFT descriptors [1 Lowe ‘99]. | ||
| Melihcan Turk F. Sukru Torun Ahmet Cagri | 10/23. | ||
| Simsek. | 11 | Dictionary Formation : Vector | |
| 2 | Outline. Introduction Bag-of-Words | Quantization. Vector quantization. 11/23. | |
| Concept Dictionary Formation Content-Based | 12 | Example Dictionary. 12/23. | |
| Image Retrieval using BoW Results | 13 | Example ‘Visual words’. 13/23. | |
| Conclusion References. 2/23. | 14 | Image Representation. codewords. | |
| 3 | Introduction : Motivation. CBIR | frequency. 14/23. | |
| motivation: Huge amount of multimedia | 15 | Content Based Image Retrieval using | |
| content demands a sophisticated analysis | BoW. We saw have to represent images using | ||
| rather than simple textual processing | the BoW concept. With histograms. It is a | ||
| (metadata such as annotations or | mapping of classical text representation | ||
| keywords). Traditional methods for | onto the image domain. Hence based on the | ||
| retrieving images is not very satisfactory | similarity of histograms we can return | ||
| or may not meet user demand E.g. In Google | ranked results, given an query image. | ||
| image typing ‘Apple’ returns the Apple | Category search: Retrieving an arbitrary | ||
| products as well as the apple fruit. Main | image representative of a specific class. | ||
| reason is the ambiguity in the language. | Used a subset of Caltech 101 dataset [2]. | ||
| Several other limitations. 3/23. | 15/23. | ||
| 4 | Introduction : Motivation. CBIR | 16 | Content Based Image Retrieval using |
| systems compensates such issues by | BoW. Given an query image return the top k | ||
| analyzing the actual ‘content’ of the | most similar results. A ‘positive’ or | ||
| image hence yielding a more effective | ‘true’ match considered to be within the | ||
| feature for describing the image rather | same category. Mean average precision | ||
| than user defined meta-data Content may be | value (MAP) is computed for each category | ||
| texture, color or any other information | using 10 query images. 16/23. | ||
| that can be derived from the image itself. | 17 | Content Based Image Retrieval using | |
| One promising idea is to represents images | BoW: Details. For vector quantization | ||
| as ‘words’ analogous to text retrieval | K-means is used with K=3000. Hence the | ||
| solutions. Document ~ Image, term (word) ~ | dictionary contains 3000 visual words and | ||
| visual word First introduces in [3]. 4/23. | the histogram has 3000 bins representing | ||
| 5 | Bag of ‘words’ Concept. 5/23. | each visual word. L2-norm – Euclidean | |
| 6 | Bag-of-Words Concept: Analogy to | distance is used for similarity measure. | |
| documents. 6/23. Of all the sensory | Visual words are represented using Lowe’s | ||
| impressions proceeding to the brain, the | SIFT descriptors. Interest points are | ||
| visual experiences are the dominant ones. | extracted using DOG (Difference of | ||
| Our perception of the world around us is | Gaussians). For each of the 18 category 10 | ||
| based essentially on the messages that | query images are used and the average MAP | ||
| reach the brain from our eyes. For a long | value is considered as the categories | ||
| time it was thought that the retinal image | success rate. 17/23. | ||
| was transmitted point by point to visual | 18 | Results : MAP of category based | |
| centers in the brain; the cerebral cortex | queries. 18/23. | ||
| was a movie screen, so to speak, upon | 19 | Results : MAP of varied dictionary | |
| which the image in the eye was projected. | sizes. 19/23. | ||
| Through the discoveries of Hubel and | 20 | Results. The ‘Motorbikes’ category has | |
| Wiesel we now know that behind the origin | the highest MAP rate (0.70). The lowest is | ||
| of the visual perception in the brain | category ‘camera’ (0.07). Average of MAP | ||
| there is a considerably more complicated | rates : 0.25 As the dictionary size get | ||
| course of events. By following the visual | larger (i.e. more visual words) images are | ||
| impulses along their path to the various | represented accurately, hence MAP values | ||
| cell layers of the optical cortex, Hubel | increase Performance seem to converge | ||
| and Wiesel have been able to demonstrate | after K>3000. 20/23. | ||
| that the message about the image falling | 21 | Conclusion. Content-Based Image | |
| on the retina undergoes a of nerve cells | Retrieval systems has gained severe | ||
| stored in columns. In this system each | interest among research scientists since | ||
| cell has its specific function and is | multimedia files such as images and videos | ||
| responsible for a specific detail in the | has dramatically entered our lives | ||
| pattern of the retinal image. | throughout the last decade Textual | ||
| 7 | Bag-of-Words Concept: Analogy to | analysis is not sufficient for effective | |
| documents. Each image can be represented | retrieval systems Analogous to document | ||
| as a histogram . where each bin of the | representation an image can be described | ||
| histogram corresponds to a visual word in | by ‘visual words’. BoW concept. Using only | ||
| the dictionary and the value of the bin is | such feature results are highly | ||
| the frequency of occurrence of such visual | satisfying. 21/23. | ||
| word. 7/23. | 22 | References. [1] D. G. Lowe. | |
| 8 | Bag of ‘words’ Concept. Hence, we | Distinctive image features from | |
| consider an image as a document. And as | scale-invariant keypoints. IJCV, | ||
| words/terms define a document, visual | 60(2):91–110, 2004 | ||
| words define an image. Words are known? | [2]http://www.vision.caltech.edu/Image_Dat | ||
| What are ‘visual words’? Need to define a | sets/Caltech101/ [3] J. Sivic and A. | ||
| dictionary. 8/23. | Zisserman. Video Google: A text retrieval | ||
| 9 | 2. 1. Bag of ‘words’ Concept : | approach to object matching in videos. In | |
| Construct a dictionary. 9/23. | Proc. ICCV, 2003. 22/23. | ||
| 10 | Dictionary Formation: Feature | 23 | Thank You! Questions and Demo! 23/23. |
| Content-Based Image Retrieval using the Bag-of-Words Concept.ppt | |||
Content-Based Image Retrieval using the Bag-of-Words Concept
«The animals» - POLAR BEAR. BEAR. WHALE. The animals which live in the rainforest and tropics. FLAMINGO. The animals which live in the OCEAN. TIGER. BOBCAT. SNAKE. The animals which live in the polar regions. DOLPHIN. REINDEER. FISH. SQUIRREL. SCORPIO. BISON. LION. GORILLA. SEAL. The animals which live in a SAVANNA.
«English words» - Translate into Russian. 1) FALSE — watch 2) FALSE — disco 3) TRUE 4) TRUE 5) FALSE — sightseeing 6) FALSE — friend 7) TRUE 8) FALSE — terrible 9) TRUE 10) FALSE — meal. Words. Translate into English. That is all! Match the expressions to others with similar meanings. Is it True or False? 1) Go and wotch a match 2) Go to a disko 3) Go shopping 4) Go for a picnic 5) Go sightseeng 6) Go and see a frend 7) Go to a party 8)I had a terible time 9) Go skiing 10) Go for a meel in a restaurant.
«Женщина the woman» - Пути пополнения лексической группы «женщина» в английском языке. Муж - голова, жена- шея; куда хочу- туда верчу. Бабий язык, куда ни завались, достанет. The wife is the key to the house. « Der mann»- нем. Холостому помогай боже, а женатому хозяйка поможет. A woman’s tongue wags like a lamb’s tail. Пословицы.
«The english-speaking countries» - USA. The English-speaking countries. Disneyland. Scotland. Australia. Great Britain.
«The green movement» - Their features. Management of each such branch in which head there is a chief executive, carries out national board. Several active workers managed to steal up on a raft to a platform and to chain themselves to it. The countries in which there are offices Greenpeace. Green color which is used by participants of movement as the general emblem, serves as a symbol of the nature, hope and updating.
«Деловое письмо на английском языке» - Вступление. Сообщение плохих новостей. Интерес. Письмо завершается и подписывается. Обучение написанию делового и личного письма на английском языке. Типы деловых писем. Прикрепленный файл. Правила оформления и структура письма личного характера. Учащиеся школ. Дополнительные вопросы. Комплекс объективных и субъективных причин.
