WhoSay was an American social media service and branding platform for celebrities and their fans. Founded in Los Angeles in 2010, with financing by Creative Artists Agency (CAA), Amazon.com and other investors, it is notable for allowing its users to retain ownership rights over the content that they post to their accounts, through copyright branding, and for enabling users to post content to other social media sites like Twitter, Facebook, Instagram and Tumblr simultaneously. WhoSay describes itself as a "social celebrity magazine" whose editorial team keeps its users informed about the latest celebrity and entertainment news. Clients such as Dylan McDermott and Chris Rock lauded the service for its ability to add content to multiple social network sites easily. Rock in particular has commented on its ease of use for those who are not part of a tech-savvy demographic, commenting, "It's perfect for someone that's not 25." WhoSay's competitors included theAudience, which is operated by the William Morris Endeavor. == History == WhoSay was founded in March 2010, by Steve Ellis and the Los Angeles-based talent agency Creative Artists Agency (CAA). It was financed through investments Amazon.com (who along with CAA, holds a minority stake in the company), Comcast, Greylock Partners, and High Peak Ventures. The company's main headquarters are in The New York Times Building in Manhattan, with additional headquarters in CAA's office building in the Silicon Beach area of Los Angeles, and in London. The company was founded to protect celebrities' intellectual property and enable the celebrities themselves to profit themselves from their own content through copyright branding. Its chief executive is co-founder Steve Ellis, who, after leaving Getty Images, was contacted by CAA, who were looking to resolve the issue of celebrities losing the rights to their own photos and videos when uploading them to social network sites. Ellis explained WhoSay's mission thus: "We work with people who are constantly being utilized by third parties for the wrong reasons. [The company was formed] to give celebrities and other influential people a set of tools to allow them to manage and control their presence in the digital world." In this way, WhoSay is likened by Ellis to "a People magazine by the people themselves who are in it." The company started slowly, until CAA client Tom Hanks signed onto WhoSay three months after the service's launch. The company continued to maintain a low profile for the first three years of operation, during which it accumulated a client list of 1,500 actors, musicians and artists. Clients are accepted by the service on an invitation-only basis, although they are not restricted to Creative Artists clients. Among them are Kelly Clarkson, Julia Louis-Dreyfus, Paula Patton, Kevin Spacey, Jim Carrey, John Cusack, Bill Maher, Johnny Knoxville, Chelsea Handler, Eva Longoria, Spike Lee, Enrique Iglesias and Katie Couric. Clients are not charged for the service, and are given a share of any revenue that is generated by advertisements. They are also given the ability share in the database of e-mail addresses that come with registration, in order to communicate directly with fans. Actor Dylan McDermott was introduced to WhoSay by his agent, as a way of easily posting content to Facebook, Twitter, Tumblr and even China's Tencent social network with relative ease. McDermott comments, "When you put something out there, you can hit everything at one time. It makes it easy for me." Comedian Chris Rock has commented that WhoSay is ideal for people like him have developed difficulty in keeping track of different websites as they get older, saying, "It's perfect for someone that's not 25." In September 2013 WhoSay introduced a mobile application for consumers. By October 2013, the company's website attracted 12 million monthly visitors. In July 2014 Rob Gregory left his role as president of Newsweek's The Daily Beast to become WhoSay's chief revenue officer. Among his responsibilities are developing ways to monetize WhoSay's web and mobile products, such as premium advertising strategies and brand partnerships. WhoSay does not allow consumers to create accounts, nor does it include search features, making it difficult to access a celebrity's account unless a user is directed there from one of their other social pages. According to Ellis, consumers have enough social media choices, saying, "Frankly they don't really need the services that we provide, and there are a lot of very specific features built into our service that really only benefit someone who is of a high profile." By February 2015, WhoSay had amassed 4.8 million unique users, and expanded its accounts to companies that employ celebrities for branded content. Such companies include Lexus, which partnered with the company to promote a campaign in which actress Rosario Dawson, during the lead up to the 87th Academy Awards, released five short videos on her social media accounts. The videos feature her driving through Los Angeles in preparation for the grand opening of her pop-up store, which sells Studio One Eighty Nine, a clothing line tied to her foundation promoting African culture and content. That April, WhoSay partnered with Chevrolet's #BestDayEver social media campaign for April Fool's Day, enlisting Olivia Wilde, Norman Reedus, Alec Baldwin, Ian Somerhalder, and Nikki Reed to surprise students in four U.S. classrooms as their substitute teachers. For example, Baldwin, dressed as Abraham Lincoln, surprised students in an Occidental College class on U.S. Culture and Society. Other companies that WhoSay has partnered with include KFC, JCPenney, Dunkin' Donuts and Crest. In January 2018, the website was acquired by Viacom (now Paramount Global).
VEX Robotics
VEX Robotics is one of the main robotics programs for elementary through university students, and a subset of Innovation First International. The VEX Robotics competitions and programs were overseen by the Robotics Education & Competition Foundation (RECF), until May 2026 when VEX split from the foundation. VEX Robotics Competition was named the largest robotics competition in the world by Guinness World Records. There are four leagues of VEX Robotics competitions designed for different age groups and skill levels: VEX V5 Robotics Competition (previously VEX EDR, VRC) is for middle and high school students, and is the largest competition out of the four. VEX Robotics teams have an opportunity to compete annually in the VEX V5 Robotics Competition (V5RC). VEX IQ Robotics Competition is for elementary and middle school students. VEX IQ robotics teams have an opportunity to compete annually in the VEX IQ Robotics Competition (VIQRC). VEX AI is a 'spinoff' of VEX U, for high school and college level students. The competition features no driver control periods, hence the name 'VEX AI'. VEX AI robotics teams have an opportunity to compete in the VEX AI Competition (VAIC). VEX U is a robotics competition for college and university students. The game is similar to V5RC, but traditionally with separate, more relaxed rules on the construction of their robots. In each of the four leagues, students are given a new challenge annually and must design, build, program, and drive a robot to complete the challenge as best they can. The robotics teams that consistently display exceptional mastery in all of these areas will eventually progress to the VEX Robotics World Championship. The description and rules for the season's competition are released during the world championship of the previous season. From 2021 to 2025, the VEX Robotics World Championship was held in Dallas, Texas each year in mid-April or mid-May, depending on which league the teams are competing in. St. Louis, Missouri will host the event in 2026 and 2027. == VEX V5 == VEX V5 is a STEM learning system designed by VEX Robotics and the REC Foundation to help middle and high school students develop problem-solving and computational thinking skills. It was introduced at the VEX Robotics World Championship in April 2019 as a replacement for a previous system called VEX EDR (VEX Cortex). The program utilizes the VEX V5 Construction and Control System as a standardized hardware, firmware, and software compatibility platform. Robotics teams and clubs can use the VEX V5 system to build robots to compete in the annual VEX V5 Robotics Competition. === Construction and Control System === The VEX V5 Construction and Control System is a metal-based robotics platform with machinable, bolt-together pieces that can be used to construct custom robotic mechanisms. The robot is controlled by a programmable processor known as the VEX V5 Brain. The Brain is equipped with a color LCD touchscreen, 21 hardware ports, an SD card port, a battery port, 8 legacy sensor ports, and a micro-USB programming port. Usage with a VEX V5 Radio enables wireless driving and wireless programming of the brain via the VEX V5 Controller. The controller allows wireless user input to the robot brain, and two controllers can be daisy-chained if necessary. Each controller has two hardware ports, a micro-USB port, two 2-axis joysticks, a monochrome LCD, and twelve buttons. The controller's LCD can be written wirelessly from the robot, providing users with configurable feedback from the robot brain. The VEX V5 Motors connect to the brain via the hardware ports and are equipped with an internal optical shaft encoder to provide feedback on the rotational status of the motor. The motor's speed is programmable but may also be altered by exchanging the internal gear cartridge with one of three cartridges of different gear ratios. The three cartridges are 100 rpm, 200 rpm, and 600 rpm. === VEXcode V5 === VEXcode V5 is a Scratch-based coding environment designed by VEX Robotics for programming VEX Robotics hardware, such as the VEX V5 Brain. The block-style interface makes programming simple for elementary through high-school students. VEXcode is consistent across VEX 123, GO, IQ, and V5 and can be used to program the devices from each. VEXcode allows the block programs to be viewed as equivalent C++ or programs to help more advanced students transition from blocks to text. This also allows easy interconversion between text-based and block-based programming. VEXcode also lets students code in C++, which gives the opportunity to learn basic C++, but to collect data from sensors or to move the drivetrain, VEX uses a header file. === PROS === PROS is a C/C++ programming environment for VEX V5 hardware maintained by students of Purdue University through Purdue ACM SIGBots. It provides a more bare-bones environment for more knowledgeable students that allows for an industry-applicable experience. It has a more robust API that allows for more precise control of the hardware for competition-level uses in VRC/VEX U. It is based on FreeRTOS. == VEX V5 Robotics Competition == VEX V5 Robotics Competition (V5RC) is a robotics competition for registered middle and high school teams that utilize the VEX V5 Construction and Control System. In this competition, teams design, cad, build, and program robots to compete at tournaments. At tournaments, teams participate in qualifying matches where two randomly chosen alliances of two teams each compete for the highest team ranking. Before the Elimination Rounds, the top-ranking teams choose their permanent alliance partners, starting with the highest-ranked team, and continuing until the alliance capacity for the tournament is reached. The new alliances then compete in an elimination bracket, and the tournament champions, alongside other award winners, qualify for their regional culminating event. . The current challenge is VEX V5 Robotics Competition: Override. === General rules === Middle and high school students have the same game and rules. The most general and basic rules for the VEX V5 Robotics Competition are as follows, but each year may have exceptions and/or additional constraints. Each robot is partnered with another robot in a pair called an "alliance". In any given match, each alliance competes against one other alliance. One team is designated as the red alliance, and the other as the blue alliance. No robot may exceed the dimensions of an 18-inch cube until the match has begun. No robot may contain hardware, software, material, or content that is not distributed by or explicitly allowed by VEX Robotics. The playing field consists of a 12-foot by 12-foot square of foam tiles bordered by a wall of metal-framed polycarbonate dividers. Anything outside of these border walls is considered as off of the playing field. The various field elements associated with that season's competition are arranged in a defined and reproducible manner before the start of each match. At the start of the match is a 15-second 'autonomous' period, where all four robots navigate the field based on pre-programmed instructions without driver input. After the autonomous period has ended, the 'driver control' period begins. This stage of the match consists of one minute and forty-five seconds of manual control of the robot using one or two handheld controllers utilized by the respective number of 'drivers'. The object of the match is to attain a higher score, i.e. more points, than the opposing alliance. The method by which the alliances attain these points varies significantly with each season. Throughout the match, the blue alliance is not allowed to enter the red alliance's 'protected zone' of the field, and vice versa. The designated areas of the field are often different for each season. During the autonomous period, the protected zone normally consists of half of the field where the alliance starts, whereas the driver control period rarely features a defined protected zone, as was the case for VRC Tipping Point, VRC High Stakes, and VRC Push Back. Intentionally removing game objects from the field will result in a warning, minor violation, and/or major violation (disqualification). Intentionally and repeatedly damaging any of the robots involved, either during the match or otherwise, will result in immediate disqualification. === 2025-2026 Game: Push Back === The objective of the game is to score as many blocks as possible in goals within a 15-second autonomous period, and 1:45 driver control period. Each field consists of two long goals, two center goals, four loaders, and two park zones. ==== Field Element - Goals ==== The goals may be pictured as 'bridges' above the field. Long goals can fit fifteen blocks of any color, while center goals can fit seven. Goals feature control bonuses that are always awarded to the alliance with the most blocks scored in the control zone of each goal. Center goal control zones inco
Law practice management software
Law practice management software is software designed to manage the business operations of a law firm. This can include software that manages cases, client intake, court communications, electronic discovery, time tracking, trust accounting, and billing. == Features of law practice management software == Common features of practice management software include: Case management Time tracking Document assembly Contact management Calendaring Docket management Client portal Contract Management Court Case Status Tracker Trust accounting == Examples of law practice management software == Smokeball LEAP Legal Software PracticeEvolve Dye & Durham
Lexxe
Lexxe is an internet search engine that applies Natural Language Processing in its semantic search technology. Founded in 2005 by Dr. Hong Liang Qiao, Lexxe is based in Sydney, Australia. Today, Lexxe's key focus is on sentiment search with the launch of a news sentiment search site at News & Moods (www.newsandmoods.com). Lexxe has experienced several stages of change of focus in search technology: Lexxe launched its Alpha version in 2005, featuring Natural Language question answering (i.e. users could ask questions in English to the search engine apart from keyword searches — this feature has been suspended for redevelopment since 2010). It used only algorithms to extract answers from web pages, with no question-answer pair databases prepared in advance. In 2011, Lexxe launched a beta version with a new search technology called Semantic Key. Semantic Keys enable users to query with a conceptual keyword (or a keyword with a special meaning, hence the term Semantic Key) in order to find instances under the concept, e.g. price → $5.95 or €200, color → red, yellow, white. For example, “price: a pound of apples”, “color: ferrari”. With initial 500 Semantic Keys at the Beta launch, Lexxe became the first search engine in the world to offer this unique and useful search technology to the users. The cost of building Semantic Keys was too heavy though. In 2017, Lexxe launched News & Moods (www.newsandmoods.com), an open platform for news sentiment search, a first step towards sentiment search feature for the entire Internet search in Lexxe search engine. News & Moods also comes with smartphone apps in Android and iOS.
Multi-scale approaches
The scale space representation of a signal obtained by Gaussian smoothing satisfies a number of special properties, scale-space axioms, which make it into a special form of multi-scale representation. There are, however, also other types of "multi-scale approaches" in the areas of computer vision, image processing and signal processing, in particular the notion of wavelets. The purpose of this article is to describe a few of these approaches: == Scale-space theory for one-dimensional signals == For one-dimensional signals, there exists quite a well-developed theory for continuous and discrete kernels that guarantee that new local extrema or zero-crossings cannot be created by a convolution operation. For continuous signals, it holds that all scale-space kernels can be decomposed into the following sets of primitive smoothing kernels: the Gaussian kernel : g ( x , t ) = 1 2 π t exp ( − x 2 / 2 t ) {\displaystyle g(x,t)={\frac {1}{\sqrt {2\pi t}}}\exp({-x^{2}/2t})} where t > 0 {\displaystyle t>0} , truncated exponential kernels (filters with one real pole in the s-plane): h ( x ) = exp ( − a x ) {\displaystyle h(x)=\exp({-ax})} if x ≥ 0 {\displaystyle x\geq 0} and 0 otherwise where a > 0 {\displaystyle a>0} h ( x ) = exp ( b x ) {\displaystyle h(x)=\exp({bx})} if x ≤ 0 {\displaystyle x\leq 0} and 0 otherwise where b > 0 {\displaystyle b>0} , translations, rescalings. For discrete signals, we can, up to trivial translations and rescalings, decompose any discrete scale-space kernel into the following primitive operations: the discrete Gaussian kernel T ( n , t ) = I n ( α t ) {\displaystyle T(n,t)=I_{n}(\alpha t)} where α , t > 0 {\displaystyle \alpha ,t>0} where I n {\displaystyle I_{n}} are the modified Bessel functions of integer order, generalized binomial kernels corresponding to linear smoothing of the form f o u t ( x ) = p f i n ( x ) + q f i n ( x − 1 ) {\displaystyle f_{out}(x)=pf_{in}(x)+qf_{in}(x-1)} where p , q > 0 {\displaystyle p,q>0} f o u t ( x ) = p f i n ( x ) + q f i n ( x + 1 ) {\displaystyle f_{out}(x)=pf_{in}(x)+qf_{in}(x+1)} where p , q > 0 {\displaystyle p,q>0} , first-order recursive filters corresponding to linear smoothing of the form f o u t ( x ) = f i n ( x ) + α f o u t ( x − 1 ) {\displaystyle f_{out}(x)=f_{in}(x)+\alpha f_{out}(x-1)} where α > 0 {\displaystyle \alpha >0} f o u t ( x ) = f i n ( x ) + β f o u t ( x + 1 ) {\displaystyle f_{out}(x)=f_{in}(x)+\beta f_{out}(x+1)} where β > 0 {\displaystyle \beta >0} , the one-sided Poisson kernel p ( n , t ) = e − t t n n ! {\displaystyle p(n,t)=e^{-t}{\frac {t^{n}}{n!}}} for n ≥ 0 {\displaystyle n\geq 0} where t ≥ 0 {\displaystyle t\geq 0} p ( n , t ) = e − t t − n ( − n ) ! {\displaystyle p(n,t)=e^{-t}{\frac {t^{-n}}{(-n)!}}} for n ≤ 0 {\displaystyle n\leq 0} where t ≥ 0 {\displaystyle t\geq 0} . From this classification, it is apparent that we require a continuous semi-group structure, there are only three classes of scale-space kernels with a continuous scale parameter; the Gaussian kernel which forms the scale-space of continuous signals, the discrete Gaussian kernel which forms the scale-space of discrete signals and the time-causal Poisson kernel that forms a temporal scale-space over discrete time. If we on the other hand sacrifice the continuous semi-group structure, there are more options: For discrete signals, the use of generalized binomial kernels provides a formal basis for defining the smoothing operation in a pyramid. For temporal data, the one-sided truncated exponential kernels and the first-order recursive filters provide a way to define time-causal scale-spaces that allow for efficient numerical implementation and respect causality over time without access to the future. The first-order recursive filters also provide a framework for defining recursive approximations to the Gaussian kernel that in a weaker sense preserve some of the scale-space properties.
Color normalization
Color normalization is a topic in computer vision concerned with artificial color vision and object recognition. In general, the distribution of color values in an image depends on the illumination, which may vary depending on lighting conditions, cameras, and other factors. Color normalization allows for object recognition techniques based on color to compensate for these variations. == Main concepts == === Color constancy === Color constancy is a feature of the human internal model of perception, which provides humans with the ability to assign a relatively constant color to objects even under different illumination conditions. This is helpful for object recognition as well as identification of light sources in an environment. For example, humans see an object approximately as the same color when the sun is bright or when the sun is dim. === Applications === Color normalization has been used for object recognition on color images in the field of robotics, bioinformatics and general artificial intelligence, when it is important to remove all intensity values from the image while preserving color values. One example is in case of a scene shot by a surveillance camera over the day, where it is important to remove shadows or lighting changes on same color pixels and recognize the people that passed. Another example is automated screening tools used for the detection of diabetic retinopathy as well as molecular diagnosis of cancer states, where it is important to include color information during classification. == Known issues == The main issue about certain applications of color normalization is that the result looks unnatural or too distant from the original colors. In cases where there is a subtle variation between important aspects, this can be problematic. More specifically, the side effect can be that pixels become divergent and not reflect the actual color value of the image. A way of combating this issue is to use color normalization in combination with thresholding to correctly and consistently segment a colored image. == Transformations and algorithms == There is a vast array of different transformations and algorithms for achieving color normalization and a limited list is presented here. The performance of an algorithm is dependent on the task and one algorithm which performs better than another in one task might perform worse in another (no free lunch theorem). Additionally, the choice of the algorithm depends on the preferences of the user for the end-result, e.g. they may want a more natural-looking color image. === Grey world === The grey world normalization makes the assumption that changes in the lighting spectrum can be modelled by three constant factors applied to the red, green and blue channels of color. More specifically, a change in illuminated color can be modelled as a scaling α, β and γ in the R, G and B color channels and as such the grey world algorithm is invariant to illumination color variations. Therefore, a constancy solution can be achieved by dividing each color channel by its average value as shown in the following formula: ( α R , β G , γ B ) → ( α R α n ∑ i R , β G β n ∑ i G , γ B γ n ∑ i B ) {\displaystyle \left(\alpha R,\beta G,\gamma B\right)\rightarrow \left({\frac {\alpha R}{{\frac {\alpha }{n}}\sum _{i}R}},{\frac {\beta G}{{\frac {\beta }{n}}\sum _{i}G}},{\frac {\gamma B}{{\frac {\gamma }{n}}\sum _{i}B}}\right)} As mentioned above, grey world color normalization is invariant to illuminated color variations α, β and γ, however it has one important problem: it does not account for all variations of illumination intensity and it is not dynamic; when new objects appear in the scene it fails. To solve this problem there are several variants of the grey world algorithm. Additionally there is an iterative variation of the grey world normalization, however it was not found to perform significantly better. === Histogram equalization === Histogram equalization is a non-linear transform which maintains pixel rank and is capable of normalizing for any monotonically increasing color transform function. It is considered to be a more powerful normalization transformation than the grey world method. The results of histogram equalization tend to have an exaggerated blue channel and look unnatural, due to the fact that in most images the distribution of the pixel values is usually more similar to a Gaussian distribution, rather than uniform. === Histogram specification === Histogram specification transforms the red, green and blue histograms to match the shapes of three specific histograms, rather than simply equalizing them. It refers to a class of image transforms which aims to obtain images of which the histograms have a desired shape. As specified, firstly it is necessary to convert the image so that it has a particular histogram. Assume an image x. The following formula is the equalization transform of this image: y = f ( x ) = ∫ 0 x p x ( u ) d u {\displaystyle y=f(x)=\int \limits _{0}^{x}p_{x}(u)du} Then assume wanted image z. The equalization transform of this image is: y ′ = g ( z ) = ∫ 0 z p z ( u ) d u {\displaystyle y'=g(z)=\int \limits _{0}^{z}p_{z}(u)du} Of course p z ( u ) {\displaystyle p_{z}(u)} is the histogram of the output image. The formula to find the inverse of the above transform is: z = g − 1 ( y ′ ) {\displaystyle z=g^{-1}(y')} Therefore, since images y and y' have the same equalized histogram they are actually the same image, meaning y = y' and the transform from the given image x to the wanted image z is: z = g − 1 ( y ′ ) = g − 1 ( y ) = g − 1 ( f ( x ) ) {\displaystyle z=g^{-1}(y')=g^{-1}(y)=g^{-1}(f(x))} Histogram specification has the advantage of producing more realistic looking images, as it does not exaggerate the blue channel like histogram equalization. === Comprehensive Color Normalization === The comprehensive color normalization is shown to increase localization and object classification results in combination with color indexing. It is an iterative algorithm which works in two stages. The first stage is to use the red, green and blue color space with the intensity normalized, to normalize each pixel. The second stage is to normalize each color channel separately, so that the sum of the color components is equal to one third of the number of pixels. The iterations continue until convergence, meaning no additional changes. Formally: Normalize the color image f ( t ) = [ f i j ( t ) ] i = 1... N , j = 1... M {\displaystyle f^{(t)}=[f_{ij}^{(t)}]_{i=1...N,j=1...M}} which consists of color vectors f i j ( t ) = ( r i j ( t ) , g i j ( t ) , b i j ( t ) ) T . {\displaystyle f_{ij}^{(t)}=(r_{ij}^{(t)},g_{ij}^{(t)},b_{ij}^{(t)})^{T}.} For the first step explained above, compute: S i j := r i j ( t ) + g i j ( t ) + b i j ( t ) {\displaystyle S_{ij}:=r_{ij}^{(t)}+g_{ij}^{(t)}+b_{ij}^{(t)}} which leads to r i j ( t + 1 ) = r i j ( t ) S i j , g i j ( t + 1 ) = g i j ( t ) S i j {\displaystyle r_{ij}^{(t+1)}={\frac {r_{ij}^{(t)}}{S_{ij}}},g_{ij}^{(t+1)}={\frac {g_{ij}^{(t)}}{S_{ij}}}} and b i j ( t + 1 ) = b i j ( t ) S i j . {\displaystyle b_{ij}^{(t+1)}={\frac {b_{ij}^{(t)}}{S_{ij}}}.} For the second step explained above, compute: r ′ = 3 N M ∑ i = 1 N ∑ j = 1 M r i j ( t + 1 ) {\displaystyle r'={\frac {3}{NM}}\sum _{i=1}^{N}\sum _{j=1}^{M}r_{ij}^{(t+1)}} and normalize r i j ( t + 2 ) = r i j ( t + 1 ) r ′ . {\displaystyle r_{ij}^{(t+2)}={\frac {r_{ij}^{(t+1)}}{r'}}.} Of course the same process is done for b' and g'. Then these two steps are repeated until the changes between iteration t and t+2 are less than some set threshold. Comprehensive color normalization, just like the histogram equalization method previously mentioned, produces results that may look less natural due to the reduction in the number of color values.
Neural style transfer
Neural style transfer (NST) software algorithms are able to manipulate digital images, or videos, in order to adopt the appearance or visual style of another image. NST algorithms are characterized by their use of deep neural networks for the sake of image transformation. Common uses for NST are the creation of artificial artwork from photographs, for example by transferring the appearance of famous paintings to user-supplied photographs. Several notable mobile apps use NST techniques for this purpose, including DeepArt and Prisma. This method has been used by artists and designers around the globe to develop new artwork based on existent style(s). == History == NST is an example of image stylization, a problem studied for over two decades within the field of non-photorealistic rendering. The first two example-based style transfer algorithms were image analogies and image quilting. Both of these methods were based on patch-based texture synthesis algorithms. Given a training pair of images–a photo and an artwork depicting that photo–a transformation could be learned and then applied to create new artwork from a new photo, by analogy. If no training photo was available, it would need to be produced by processing the input artwork; image quilting did not require this processing step, though it was demonstrated on only one style. NST was first published in the paper "A Neural Algorithm of Artistic Style" by Leon Gatys et al., originally released to ArXiv 2015, and subsequently accepted by the peer-reviewed CVPR conference in 2016. The original paper used a VGG-19 architecture that has been pre-trained to perform object recognition using the ImageNet dataset. In 2017, Google AI introduced a method that allows a single deep convolutional style transfer network to learn multiple styles at the same time. This algorithm permits style interpolation in real-time, even when done on video media. == Mathematics == This section closely follows the original paper. === Overview === The idea of Neural Style Transfer (NST) is to take two images—a content image p → {\displaystyle {\vec {p}}} and a style image a → {\displaystyle {\vec {a}}} —and generate a third image x → {\displaystyle {\vec {x}}} that minimizes a weighted combination of two loss functions: a content loss L content ( p → , x → ) {\displaystyle {\mathcal {L}}_{\text{content }}({\vec {p}},{\vec {x}})} and a style loss L style ( a → , x → ) {\displaystyle {\mathcal {L}}_{\text{style }}({\vec {a}},{\vec {x}})} . The total loss is a linear sum of the two: L NST ( p → , a → , x → ) = α L content ( p → , x → ) + β L style ( a → , x → ) {\displaystyle {\mathcal {L}}_{\text{NST}}({\vec {p}},{\vec {a}},{\vec {x}})=\alpha {\mathcal {L}}_{\text{content}}({\vec {p}},{\vec {x}})+\beta {\mathcal {L}}_{\text{style}}({\vec {a}},{\vec {x}})} By jointly minimizing the content and style losses, NST generates an image that blends the content of the content image with the style of the style image. Both the content loss and the style loss measures the similarity of two images. The content similarity is the weighted sum of squared-differences between the neural activations of a single convolutional neural network (CNN) on two images. The style similarity is the weighted sum of Gram matrices within each layer (see below for details). The original paper used a VGG-19 CNN, but the method works for any CNN. === Symbols === Let x → {\textstyle {\vec {x}}} be an image input to a CNN. Let F l ∈ R N l × M l {\textstyle F^{l}\in \mathbb {R} ^{N_{l}\times M_{l}}} be the matrix of filter responses in layer l {\textstyle l} to the image x → {\textstyle {\vec {x}}} , where: N l {\textstyle N_{l}} is the number of filters in layer l {\textstyle l} ; M l {\textstyle M_{l}} is the height times the width (i.e. number of pixels) of each filter in layer l {\textstyle l} ; F i j l ( x → ) {\textstyle F_{ij}^{l}({\vec {x}})} is the activation of the i th {\textstyle i^{\text{th}}} filter at position j {\textstyle j} in layer l {\textstyle l} . A given input image x → {\textstyle {\vec {x}}} is encoded in each layer of the CNN by the filter responses to that image, with higher layers encoding more global features, but losing details on local features. === Content loss === Let p → {\textstyle {\vec {p}}} be an original image. Let x → {\textstyle {\vec {x}}} be an image that is generated to match the content of p → {\textstyle {\vec {p}}} . Let P l {\textstyle P^{l}} be the matrix of filter responses in layer l {\textstyle l} to the image p → {\textstyle {\vec {p}}} . The content loss is defined as the squared-error loss between the feature representations of the generated image and the content image at a chosen layer l {\displaystyle l} of a CNN: L content ( p → , x → , l ) = 1 2 ∑ i , j ( A i j l ( x → ) − A i j l ( p → ) ) 2 {\displaystyle {\mathcal {L}}_{\text{content }}({\vec {p}},{\vec {x}},l)={\frac {1}{2}}\sum _{i,j}\left(A_{ij}^{l}({\vec {x}})-A_{ij}^{l}({\vec {p}})\right)^{2}} where A i j l ( x → ) {\displaystyle A_{ij}^{l}({\vec {x}})} and A i j l ( p → ) {\displaystyle A_{ij}^{l}({\vec {p}})} are the activations of the i th {\displaystyle i^{\text{th}}} filter at position j {\displaystyle j} in layer l {\displaystyle l} for the generated and content images, respectively. Minimizing this loss encourages the generated image to have similar content to the content image, as captured by the feature activations in the chosen layer. The total content loss is a linear sum of the content losses of each layer: L content ( p → , x → ) = ∑ l v l L content ( p → , x → , l ) {\displaystyle {\mathcal {L}}_{\text{content }}({\vec {p}},{\vec {x}})=\sum _{l}v_{l}{\mathcal {L}}_{\text{content }}({\vec {p}},{\vec {x}},l)} , where the v l {\displaystyle v_{l}} are positive real numbers chosen as hyperparameters. === Style loss === The style loss is based on the Gram matrices of the generated and style images, which capture the correlations between different filter responses at different layers of the CNN: L style ( a → , x → ) = ∑ l = 0 L w l E l , {\displaystyle {\mathcal {L}}_{\text{style }}({\vec {a}},{\vec {x}})=\sum _{l=0}^{L}w_{l}E_{l},} where E l = 1 4 N l 2 M l 2 ∑ i , j ( G i j l ( x → ) − G i j l ( a → ) ) 2 . {\displaystyle E_{l}={\frac {1}{4N_{l}^{2}M_{l}^{2}}}\sum _{i,j}\left(G_{ij}^{l}({\vec {x}})-G_{ij}^{l}({\vec {a}})\right)^{2}.} Here, G i j l ( x → ) {\displaystyle G_{ij}^{l}({\vec {x}})} and G i j l ( a → ) {\displaystyle G_{ij}^{l}({\vec {a}})} are the entries of the Gram matrices for the generated and style images at layer l {\displaystyle l} . Explicitly, G i j l ( x → ) = ∑ k F i k l ( x → ) F j k l ( x → ) {\displaystyle G_{ij}^{l}({\vec {x}})=\sum _{k}F_{ik}^{l}({\vec {x}})F_{jk}^{l}({\vec {x}})} Minimizing this loss encourages the generated image to have similar style characteristics to the style image, as captured by the correlations between feature responses in each layer. The idea is that activation pattern correlations between filters in a single layer captures the "style" on the order of the receptive fields at that layer. Similarly to the previous case, the w l {\displaystyle w_{l}} are positive real numbers chosen as hyperparameters. === Hyperparameters === In the original paper, they used a particular choice of hyperparameters. The style loss is computed by w l = 0.2 {\displaystyle w_{l}=0.2} for the outputs of layers conv1_1, conv2_1, conv3_1, conv4_1, conv5_1 in the VGG-19 network, and zero otherwise. The content loss is computed by w l = 1 {\displaystyle w_{l}=1} for conv4_2, and zero otherwise. The ratio α / β ∈ [ 5 , 50 ] × 10 − 4 {\displaystyle \alpha /\beta \in [5,50]\times 10^{-4}} . === Training === Image x → {\displaystyle {\vec {x}}} is initially approximated by adding a small amount of white noise to input image p → {\displaystyle {\vec {p}}} and feeding it through the CNN. Then we successively backpropagate this loss through the network with the CNN weights fixed in order to update the pixels of x → {\displaystyle {\vec {x}}} . After several thousand epochs of training, an x → {\displaystyle {\vec {x}}} (hopefully) emerges that matches the style of a → {\displaystyle {\vec {a}}} and the content of p → {\displaystyle {\vec {p}}} . As of 2017, when implemented on a GPU, it takes a few minutes to converge. == Extensions == In some practical implementations, it is noted that the resulting image has too much high-frequency artifact, which can be suppressed by adding the total variation to the total loss. Compared to VGGNet, AlexNet does not work well for neural style transfer. NST has also been extended to videos. Subsequent work improved the speed of NST for images by using special-purpose normalizations. In a paper by Fei-Fei Li et al. adopted a different regularized loss metric and accelerated method for training to produce results in real-time (three orders of magnitude faster than Gatys). Their idea was to use not the pixel-based loss defined above but rather a 'perceptual loss' measuring t