25:null 26:["$","title",null,{"children":"Final Project | ychsu6 | Observable"}] 2b:T429,md` ## Overview Cubism has its foundation in forcing multiple perspectives simultaneously and decomposing realistic subjects, to give them perspective and distinct impressions. Therefore, the interpreter here plays an important role as it determines how the viewer’s perceptions are shaped, thus giving the artist power over the subject. However, in the history of art, such power relationships very often remain male-dominated, due to the predominant male artists of its lead names and the female Muses who create nothing but put themselves at the service of Masters, leading to the idea that females are merely described as inspirations behind artworks and female artists’ accomplishments are ignored. Therefore, this project aims to reverse the system by taking faceted components of individual parts of “attractive” “male” celebrities (as Muses); arbitrarily forcing multiple perspectives and recombining them together to put into a scary, indistinguishable, not unattractive monster; then applying with female cubists’ (as Masters) styles. `2c:T493,md` ## Methodology & Techinical ### Deconstucted Attrictive Male Facial Components - CelebAMask-HQ Dataset images (30,000 faces selected by CelebA dataset) & CelebAMask-HQ Dataset Mask annotations (B& W threshold grayscale) *Dataset-> Female: Male = 2:1 - 40* (1, -1 ) attributes CelebA dataset (300,000 faces) - Position annotation (x, y, z) of faces for CelebAMask-HQ Dataset images

Apply CelebAMask-HQ Dataset Mask annotations as alpha value to CelebAMask-HQ Dataset images to return the separated components images.
Approach: Convert image and reSized components mask to PNG by running script in Terminal with [Imagemagick](https://imagemagick.org/index.php) dependency + trim: convert main.jpg mask.jpg -resize 1024x1024 -alpha off -compose copy-opacity -composite -trim +repage new.png

`2d:T7c7,md` ### Faceted Facial Template & Corresponding MidPoints and Scale

Each facial component can be shown with array of data:
[Mid-point of X-coordinate, Mid-point of Y-coordinate, width]
[hairL, hairR, faceL, faceR, browL, browR, eyeL, eyeR, noseL, noseR, lipRU, lipRL, lipLU, lipLL, mouthR]

Eventually the mouth component had to be removed since mouth annotations are missing in quite some images. (Due to some technical difficulties and time limit, eventually, only template 1 is fetched for the demo work. Here the template (contains female parts) only acts as visual references for the facial composition, but all components are selected from males in the demo.) ### Female Cubists' Styles Transfered - [Marevna Marie Vorobieff](https://www.wikiart.org/en/marevna-marie-vorobieff/all-works#!#filterName:Style_cubism,resultType:masonry) - [Natalia Goncharova](https://www.wikiart.org/en/natalia-goncharova/all-works#!#filterName:Style_cubism,resultType:masonry) - [Marie Laurencin](https://www.wikiart.org/en/marie-laurencin/all-works#!#filterName:Style_cubism,resultType:masonry) - [Tamara de Lempicka](https://www.wikiart.org/en/tamara-de-lempicka/all-works#!#filterName:Style_cubism,resultType:masonry) - [Lyubov Popova](https://www.wikiart.org/en/lyubov-popova/all-works#!#filterName:Style_cubism,resultType:masonry) - [Olga Rozanova](https://www.wikiart.org/en/olga-rozanova/all-works#!#filterName:Style_cubism,resultType:masonry)
#### 40 Cubism portrait works by Female Cubists as Styles

`2e:T7df,md` ## Why these tools? ### Why CelebA Dataset? [CelebA Dataset](https://www.kaggle.com/jessicali9530/celeba-dataset) (Celebrities data with 40 binary facial attributes) is clearly biased, by including not only subjective facial attributes but also objective classifications that not even humans can distinguish, or attributes that are inherently not supposed to be binarised. Meanwhile, celebrities are those who are well-known by the general public (for all possible reasons), especially in the entertainment business. However, by merely selecting one of the most bias attributes, “attractive”, with the “male” attribute in the dataset, these realistic components of “Muses” are being objectified and decomposed. ### Why Style Transfer? After getting the first blended version (all generated style looks quite similar) from smaller models, I was wondering whether Style Transfer was just "generalizing", giving an overview, or merely "fetching" color palette and randomly assigning onto images, because strokes, shapes and compositions, (all that make Cubism works different from one and another) are gone. Therefore, I spend a few days digging a little bit more into the algorithm, trying to question whether such style transfer models are sufficient to work as real artists' jobs to generate a "representation" for the subject. After going through Dumoulin, et. al.’s and Ghiasi, et. al.'s papers, I realized how these losses are extracted (the distribution of brush strokes is taken out at some of the early layers while the identification of content is extracted at a much deeper layer), it does match with how I wish the style can be transferred in my work. Plus, I figured that just by simply replacing both the Style Model and the Transformer Model from small ones to big ones, the visual results look significantly better!
More Info on Process Documentation Notebook for [how I tested out and research on Style Transfer](https://observablehq.com/@betahsu/final-project/2#cell-740) `2f:T662,md` ## Aesthetical Considerations ### Style Transfer Models and Parameters As included in the Implementation website for an [arbitrary style transfer algorithm](https://reiinakano.com/arbitrary-image-stylization-tfjs/) (running purely in the browser using TensorFlow.js), there are several parameters one may play around with:
Content Image Size, Style Image Size, Style Ratio, and 4 Models (2 Style Models and 2 Transformer Models), which are:
[High Quality] Original Inceptionv3 style model (36.3MB)
[Fast] Distilled MobileNet style model (9.6MB)
[High Quality] Original transformer model (7.9MB)
[Fast] Separable_conv2d transformer (2.4MB)
In general, a smaller style image and a larger style ratio give out relatively more stylized results. And of course, larger models (take longer to load) produce more accurate style-transferred results.

Smaller transformer provides more Generalized outputs, while more distinct style differences can be seen in in outputs from larger transformer model.
More Info on Process Documentation Notebook for [more testing visual results](https://observablehq.com/@betahsu/final-project/2#cell-513) In this project, Separable_conv2d transformer (bigger transformer model) and Distilled MobileNet style model (smaller style model) are selected, as the larger transformer distinctly provides better results in transforming styles, while the smaller style model creates interesting textures and strokes which I wish to preserve in the work. `30:[{"id":"9956c888b5793f2e88344bc91f390ecfcb9a52ff62320dcd9ac7251e5a3c7cbfa8434d489b42e9c981b9d22c35a32e1a8700786be0ae985364e1bb9874775d94","href":"https://static.observableusercontent.com/files/9956c888b5793f2e88344bc91f390ecfcb9a52ff62320dcd9ac7251e5a3c7cbfa8434d489b42e9c981b9d22c35a32e1a8700786be0ae985364e1bb9874775d94","name":"tryOut-5.png","type":"image/png","lastModified":1639923255964.819,"size":715774},{"id":"43111985e161ca3a124eb5f583bf83f6c88e8567e6b3bec7d3aedadb73a46bcda7a96f9d51a61456303ed5c7f0561e6fe16103386cdbe39de25bab267dbb7446","href":"https://static.observableusercontent.com/files/43111985e161ca3a124eb5f583bf83f6c88e8567e6b3bec7d3aedadb73a46bcda7a96f9d51a61456303ed5c7f0561e6fe16103386cdbe39de25bab267dbb7446","name":"304.jpg","type":"image/jpeg","lastModified":1639922889080.98,"size":104122},{"id":"2a74f07cfffe11bc8a90e7f03cafd561ae403f4203451e60df91629b2d3b47d82b35ad3be70d53b15d1cb33fae4b0ef8d248b3f7cd48882ec82569df4b892f14","href":"https://static.observableusercontent.com/files/2a74f07cfffe11bc8a90e7f03cafd561ae403f4203451e60df91629b2d3b47d82b35ad3be70d53b15d1cb33fae4b0ef8d248b3f7cd48882ec82569df4b892f14","name":"tryOut-4.png","type":"image/png","lastModified":1639885534865.636,"size":636477},{"id":"8449b8ca1571de4bc98b9dd1a09e50bc465a4c373230aa84cf8d484edbe431d137871cbc83513395fa70147a13af284febff23514dad8a3574d3b6a4fa61a925","href":"https://static.observableusercontent.com/files/8449b8ca1571de4bc98b9dd1a09e50bc465a4c373230aa84cf8d484edbe431d137871cbc83513395fa70147a13af284febff23514dad8a3574d3b6a4fa61a925","name":"tryOut-2(1).png","type":"image/png","lastModified":1639885542644.883,"size":675207},{"id":"1b4e3cd88739c8790acd6d2e75c9f5c3a0e054e38503fe92dbaf11e784cb6387f81cd93fc6f4f7990cae47cacb3678b8794380894408792c5c8419c9b6ec5b65","href":"https://static.observableusercontent.com/files/1b4e3cd88739c8790acd6d2e75c9f5c3a0e054e38503fe92dbaf11e784cb6387f81cd93fc6f4f7990cae47cacb3678b8794380894408792c5c8419c9b6ec5b65","name":"outcome-2.png","type":"image/png","lastModified":1639954162933.442,"size":1550209},{"id":"5788bc6d36b14b7162b2a8fe6eb67f4400739942345d38906db113c712622eb3b90146a6b6ec8a5cf8a8f74ea9631f897f0eddc2c528e199e318a3db5dcdb783","href":"https://static.observableusercontent.com/files/5788bc6d36b14b7162b2a8fe6eb67f4400739942345d38906db113c712622eb3b90146a6b6ec8a5cf8a8f74ea9631f897f0eddc2c528e199e318a3db5dcdb783","name":"outcome-3.png","type":"image/png","lastModified":1639954363528.78,"size":1523744},{"id":"f0857efb894130d546395467ed160208b50a1f18c66310416f0a2e9658b6bd4ab66c22728b68657fbdf6e7ce38bdf70150c0ae8107e5607baad7173f4565604e","href":"https://static.observableusercontent.com/files/f0857efb894130d546395467ed160208b50a1f18c66310416f0a2e9658b6bd4ab66c22728b68657fbdf6e7ce38bdf70150c0ae8107e5607baad7173f4565604e","name":"outcome-1.png","type":"image/png","lastModified":1639954160968.311,"size":1719183},{"id":"4442129948d587afa08b467b841f1029aace50270e97b773dee5982d8b07b221d401627a7f8cb84b9dfc67c85f2a589761d0a116d67b407523183580cd9ee663","href":"https://static.observableusercontent.com/files/4442129948d587afa08b467b841f1029aace50270e97b773dee5982d8b07b221d401627a7f8cb84b9dfc67c85f2a589761d0a116d67b407523183580cd9ee663","name":"outcome.png","type":"image/png","lastModified":1639954155496.794,"size":1405057},{"id":"27c4b64cbea787bb646c6c7beb5c76309cb8892d9be99d469ae0fc29a2bdd84ff088c65259bed852988912bd08742c9deb3f9e044f5c15a85badf327f2d27bbd","href":"https://static.observableusercontent.com/files/27c4b64cbea787bb646c6c7beb5c76309cb8892d9be99d469ae0fc29a2bdd84ff088c65259bed852988912bd08742c9deb3f9e044f5c15a85badf327f2d27bbd","name":"00304_nose@1.png","type":"image/png","lastModified":1639922947331.867,"size":55441},{"id":"8f7b5ab580e691cc0cc0c0b4ebda91b0da075a92c7b437dcaed62f7bbf1c0a7fd6e345a2832cac510adb15baab5738de8c9a3b3b79002d61a6754062c9422b44","href":"https://static.observableusercontent.com/files/8f7b5ab580e691cc0cc0c0b4ebda91b0da075a92c7b437dcaed62f7bbf1c0a7fd6e345a2832cac510adb15baab5738de8c9a3b3b79002d61a6754062c9422b44","name":"00304_nose.png","type":"image/png","lastModified":1639922920614.508,"size":1415},{"id":"a696dfa6142596ec7e9433814111bf745be574f9d22f1d40266a296608ffb3618730831c303ed4d76c94a3ad178bc595178ea79b978a7d745a774df2e115b0c2","href":"https://static.observableusercontent.com/files/a696dfa6142596ec7e9433814111bf745be574f9d22f1d40266a296608ffb3618730831c303ed4d76c94a3ad178bc595178ea79b978a7d745a774df2e115b0c2","name":"outcome-4.png","type":"image/png","lastModified":1639958653860.6,"size":1565979},{"id":"1b87c31ad683aa8c2aefeceac46ad87a72ab79acd39c6b3f01f80b3ac2deb7cb78a9e0d3176720dbadb6710e51ea6a118d1cda746ab08e70077ea7c0de53ec6c","href":"https://static.observableusercontent.com/files/1b87c31ad683aa8c2aefeceac46ad87a72ab79acd39c6b3f01f80b3ac2deb7cb78a9e0d3176720dbadb6710e51ea6a118d1cda746ab08e70077ea7c0de53ec6c","name":"outcome-5.png","type":"image/png","lastModified":1639958653647.7,"size":1115318},{"id":"4e5d66a7c5bae9d3e18f44971c8e1677b5a3650f4cab01a98b03a7fb238d63a265dc682f499e63a48bbd88c9985de777415a93d8ee650f3cb80ddcf55ef5d984","href":"https://static.observableusercontent.com/files/4e5d66a7c5bae9d3e18f44971c8e1677b5a3650f4cab01a98b03a7fb238d63a265dc682f499e63a48bbd88c9985de777415a93d8ee650f3cb80ddcf55ef5d984","name":"Screenshot 2021-12-20 at 7.59.18 AM.png","type":"image/png","lastModified":1639958374211.746,"size":2581489},{"id":"7c9aae93730dc7bbfe98c8d9f3e56bee0611bb2a9f8cd40da1ca84a366fc0fea0e63606563ac9e7b285c908aba851324bfed5b6f77807b3438ab71c5e1b1be3b","href":"https://static.observableusercontent.com/files/7c9aae93730dc7bbfe98c8d9f3e56bee0611bb2a9f8cd40da1ca84a366fc0fea0e63606563ac9e7b285c908aba851324bfed5b6f77807b3438ab71c5e1b1be3b","name":"faces.png","type":"image/png","lastModified":1639963646075.228,"size":1597655},{"id":"60a0da01fe253d904ef51e169205b4d31b096d7722050fbad8017e6a36fe054481e0d8d1660feec08cb0488f98371365cb04bf46cf3a2c29228eeccc64942b11","href":"https://static.observableusercontent.com/files/60a0da01fe253d904ef51e169205b4d31b096d7722050fbad8017e6a36fe054481e0d8d1660feec08cb0488f98371365cb04bf46cf3a2c29228eeccc64942b11","name":"attractiveMales.png","type":"image/png","lastModified":1639964908616.887,"size":87272},{"id":"e938ddd6d97cf9e84ffc27ffa0613263f68eb6bfc4655397fb08e0083ba36086b8d186d836c949c3b359f55d1252570e5bf8e7e2812232eb6a36bd52e5d02c04","href":"https://static.observableusercontent.com/files/e938ddd6d97cf9e84ffc27ffa0613263f68eb6bfc4655397fb08e0083ba36086b8d186d836c949c3b359f55d1252570e5bf8e7e2812232eb6a36bd52e5d02c04","name":"pose@1.png","type":"image/png","lastModified":1639964904615.689,"size":63178},{"id":"813850486c4043de0d15175bb0ff0f551d86612781c6bafac09d1ea9294708f8e769cd9cc3c0ba548c76f3d2420edce6eb234a575ed702cf65b213dcfa7f27cf","href":"https://static.observableusercontent.com/files/813850486c4043de0d15175bb0ff0f551d86612781c6bafac09d1ea9294708f8e769cd9cc3c0ba548c76f3d2420edce6eb234a575ed702cf65b213dcfa7f27cf","name":"outcome-7.png","type":"image/png","lastModified":1639969176417.841,"size":1434703},{"id":"f7eb73398f9f48cd1fc7110b75d6a9f38452faec1c7a67ab812ad143dded5294868991e1b8eb739624147a703aef100b93e774f633d5283bed4ff52f225ff19d","href":"https://static.observableusercontent.com/files/f7eb73398f9f48cd1fc7110b75d6a9f38452faec1c7a67ab812ad143dded5294868991e1b8eb739624147a703aef100b93e774f633d5283bed4ff52f225ff19d","name":"outcome-6.png","type":"image/png","lastModified":1639969101011.278,"size":1240606},{"id":"6bb7df2bdc24e8872053c8ff6b91bd2d0d9a0873d748207bc30c2556916ffad9f2e90287d04ca2e778d64e56509764cd56f9b6532ca1454aecede5d5db0996f3","href":"https://static.observableusercontent.com/files/6bb7df2bdc24e8872053c8ff6b91bd2d0d9a0873d748207bc30c2556916ffad9f2e90287d04ca2e778d64e56509764cd56f9b6532ca1454aecede5d5db0996f3","name":"download (1).png","type":"image/png","lastModified":1639968991542.431,"size":1461246},{"id":"4de246f6b1d16d06b2bb5576019a59a7d5968bf43d4f78a362d55aca1d062890d52473c04d9fa28fdf6f511c3b2845a7daba690f4a5213a8303dfca46eba34cb","href":"https://static.observableusercontent.com/files/4de246f6b1d16d06b2bb5576019a59a7d5968bf43d4f78a362d55aca1d062890d52473c04d9fa28fdf6f511c3b2845a7daba690f4a5213a8303dfca46eba34cb","name":"videoWalkThrough_1.mp4","type":"video/mp4","lastModified":1639968765366.402,"size":34458454},{"id":"c29bf2ad2f3a897ace85967ebe64d6955d35778c184b5a4fbd6a141f6abf9fd9bbc3ba323fb5b6eb198e210a4cb138d052875c096f6b4f3d0bd7bbfc4fa38e14","href":"https://static.observableusercontent.com/files/c29bf2ad2f3a897ace85967ebe64d6955d35778c184b5a4fbd6a141f6abf9fd9bbc3ba323fb5b6eb198e210a4cb138d052875c096f6b4f3d0bd7bbfc4fa38e14","name":"Screenshot 2021-12-20 at 12.26.10 PM.png","type":"image/png","lastModified":1639974379523.715,"size":328045},{"id":"28a5bb35e38c16700506a596991e304e88e72e99e628e101f4336d7e5ee2555193311dbd0163282379b2b53fd98da1d3e37be179511b9e26bdb9c5c897043a27","href":"https://static.observableusercontent.com/files/28a5bb35e38c16700506a596991e304e88e72e99e628e101f4336d7e5ee2555193311dbd0163282379b2b53fd98da1d3e37be179511b9e26bdb9c5c897043a27","name":"Screenshot 2021-12-20 at 12.03.29 PM.png","type":"image/png","lastModified":1639973017819.188,"size":135643},{"id":"cb805dcabfac26247a35981478a4cfa966a3401af1d338c81f1f2363e9fdcf78116cd6079628c97903024f24e80f12233a3d0d42a826c52f7745ca7697923867","href":"https://static.observableusercontent.com/files/cb805dcabfac26247a35981478a4cfa966a3401af1d338c81f1f2363e9fdcf78116cd6079628c97903024f24e80f12233a3d0d42a826c52f7745ca7697923867","name":"Screenshot 2021-12-21 at 4.24.27 AM.png","type":"image/png","lastModified":1640031909185.799,"size":1228742}] 29:{"id":"d5994a6744006b81","slug":"final-project/3","title":"Final Project","thumbnail":"47273f3f1f9c7955fded0ef9bf814a06a041251aa66269190fcfc9cfeadec403","access_level":"public","create_time":"$D2021-12-18T15:17:17.612Z","update_time":"$D2021-12-20T20:29:49.491Z","view_time":null,"first_public_time":"$D2021-12-18T15:18:05.305Z","has_unpublished_edits":null,"fork":false,"stars":2,"starred":false,"owner":{"type":"team","id":"c3d91d7db2867b4d","login":"betahsu","name":"ychsu6","avatar_url":"https://avatars-next.observableusercontent.com/197b7792c36f626b9fe391a472486b8b6dc3cf3503a2bb52e365e7916c324a9e","tier":"starter_2024"},"version":311,"publish_version":311,"role":null,"authored":false,"authors":[{"type":"individual","id":"8205ffe96d046f8c","login":"betahsu","name":"ychsu6","avatar_url":"https://avatars-next.observableusercontent.com/197b7792c36f626b9fe391a472486b8b6dc3cf3503a2bb52e365e7916c324a9e","tier":null}],"members":[{"type":"individual","id":"8205ffe96d046f8c","login":"betahsu","name":"ychsu6","avatar_url":"https://avatars-next.observableusercontent.com/197b7792c36f626b9fe391a472486b8b6dc3cf3503a2bb52e365e7916c324a9e","tier":null,"role":"editor"}],"comments":[],"custom_thumbnail":true,"trashed":false,"lapsed":null,"subscribed":null,"forks":0,"collection":null,"type":"notebook","body":{"title":"Final Project","cells":[{"id":0,"value":"# Final Project","pinned":false,"hidden":false,"mode":"md","author":"$undefined","summary":"$undefined","output":"","database":"$undefined"},{"id":78,"value":"md`\n## Her Muse
\nBeta HSU Yun Chu, 2021
\nMedium: CelebADataset, StyleTransfer, Tensorflow.js, P5js, Imagemagick
\nTopic: #styleTransfer, #neuralNetwork, # generativeArt, #critical, #muse, #femaleCubists, #maleCelebrities\n`","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","output":"$undefined","database":"$undefined"},{"id":259,"value":"md`\n## [Final Outcomes](https://ppegy.csb.app/)\n(Click to run in the browser, refresh the page to generate the new)
\nCoded in CodeSandBox\n`","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","output":"$undefined","database":"$undefined"},{"id":120,"value":"md`\n## ScreenShots\n

\n\n

\n`","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","output":"$undefined","database":"$undefined"},{"id":31,"value":"$2b","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","output":"$undefined","database":"$undefined"},{"id":238,"value":"md`\n## Video demo\n

\n`","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","output":"$undefined","database":"$undefined"},{"id":57,"value":"$2c","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","output":"$undefined","database":"$undefined"},{"id":111,"value":"$2d","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","output":"$undefined","database":"$undefined"},{"id":204,"value":"md`\n### Filter Out Attractive + Male Celibrities & Categorize pose data\n#### Filtering Attractive Males\n

\n#### Categorize pose data into 3 groups, via X coordinates value\n

\n### Fetching images and applying via filtering\n

\n### Apply Style Transfer via Tensorflow.js\nTensorFlow and P5.js utilities [Demo and Testing Code]((https://mlq4l.csb.app/)\n\n`","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","output":"$undefined","database":"$undefined"},{"id":168,"value":"md`\n### Challenges \n#### The Dataset is too large to upload\nSince CodeSandBox couldn’t handle such a large dataset, it crashed several times and then refused to upload any more images; thus even ALL files are pre-processed/filtered in code, only very few numbers of facial components images are uploaded and recombined into the Muze, thus one might notice that some facial components might are shown repeatedly. \n`","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","output":"$undefined","database":"$undefined"},{"id":60,"value":"$2e","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","output":"$undefined","database":"$undefined"},{"id":174,"value":"$2f","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","output":"$undefined","database":"$undefined"},{"id":229,"value":"md`\n## Possible Further Improvements\nFetching such a large image dataset easily while running in the browser without having to manually upload image dataset or cutting down on numbers?
\nInteractive work that allows users inputs to control the parameters in Style Transfer?
\nAnimate the stylized process, like what it's done in my [TensorFlow and P5.js utilities Demo and Testing Code]((https://mlq4l.csb.app/)?
\n`","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","output":"$undefined","database":"$undefined"},{"id":67,"value":"md`\n## References\nCelebA Dataset: https://www.kaggle.com/jessicali9530/celeba-dataset
\nWikiArt female Cubists and their works: https://www.wikiart.org/en/artists-by-art-movement/cubism#!#resultType:masonry
\nStyle Transfer running in the browser using TensorFlow.js: https://github.com/reiinakano/arbitrary-image-stylization-tfjs
\nStyle Transfer research:
\nhttps://magenta.tensorflow.org/2016/11/01/multistyle-pastiche-generator\nhttps://magenta.tensorflow.org/blog/2018/12/20/style-transfer-js/\nhttps://ai.googleblog.com/2016/10/supercharging-style-transfer.html
\nTensorflow API: https://js.tensorflow.org/api/1.7.1
\nTensorflow.js on Geeks for geeks: https://www.geeksforgeeks.org/tensorflow-js-tf-tensor-function/?ref=gcse
\nJaveScript- Promise: https://javascript.info/promise-basics
\n`","pinned":false,"hidden":false,"mode":"ojs","author":"$undefined","summary":"$undefined","output":"$undefined","database":"$undefined"}],"files":"$W30","stdlib":"1","titling":"auto"}} 2a:"$29:body"