2d:null 2e:["$","title",null,{"children":"ML for Artist Final Project | ychsu6 | Observable"}] 33:T912,## Machine Learning techniques inspirations ### Dimensional reduction techniques - **PCA** for reducing the dimensionality of data by projecting data onto a lower-dimensional subspace that retains most of the variance among the data points, while preserving as much as possible of the information contained in the original data. Therefore, **similar data points should cluster together** because it is explicitly focusing on axes that maximize the variance. - [**Autoencoders**](https://hackernoon.com/autoencoders-deep-learning-bits-1-11731e200694?ref=hackernoon.com) are neural networks that can be used to reduce the data into a low dimensional [latent space](https://hackernoon.com/latent-space-visualization-deep-learning-bits-2-bd09a46920df) by stacking multiple non-linear transformations(layers) contains a simpler representation of the original datasets and shows much denser clusters. - For dimensionality reduction to be effective, there needs to be an underlying low-dimensional structure in the feature space
**Example:** - Plotting the t-SNE embedding of the MNIST of Handwritten digits dataset (image space v.s. latent space) in 2-D projection comparison

- Handwritten digits (image space v.s. latent space) linear interpolation comparison

- Further possibilities -- **Latent space arithmetic**: instead of interpolating, we can add or subtract latent space representations.

- Technical Articles references: - [Understanding PCA v.s. Autoencoders](https://towardsdatascience.com/understanding-pca-autoencoders-algorithms-everyone-can-understand-28ee89b570e2) - [PCA v.s. Autoencoders visual results comparison](https://towardsdatascience.com/autoencoders-vs-pca-when-to-use-which-73de063f5d7)
34:T447,## Datasets & ML models inspirations ### [Accurate 3D Body Shape Regression using Metric and Semantic Attributes](https://shapy.is.tue.mpg.de/index.html) ([research paper](https://ps.is.mpg.de/uploads_file/attachment/attachment/691/00928.pdf))
The project aims to create a model that generates accurate 3D body shape models from images (taken from model agency sites). With the use of **Anthropometric measurements** (height, waist, hips, bust) & **Linguistic shape attributes**.

**Linguistic shape attributes**: linguistic body shape annotations they collected by asking workers on Amazon Mechanical Turk how much a word, such as short or big, applies to the model's body shape. In total **15 attributes**, 6 attributes among all are gender-specific.

35:T842,## My proposed "ways of seeing" to break such stereotypical gender duality & body shape bias #### What if such gender body shape bias can be visualized?

#### Gender-specific morphing model 1 - Taking this **14-dimensional** (excluding the gender evaluation score) Linguistic Shape Attributes Dataset, reduce the dimension via PCA to get a more dense representation that emphasizes "similar body shapes" within the same gender of the dataset in space. - Revisit points in the latent space, every time always visit the closest point to get a "morphing" of body shapes. Then (very) later visualize the gradual morphing/changes of body shapes via the corresponding 3d pose data.

(here we get 2 morphing results, 1 male, 1 female)
#### Non-gender-specific morphing model 2 - Additionally, by taking this **10 dimensional** (non-gender-specific + gender evaluation score, by matching the masculine score with the reversed of the feminine score) reduce the dimension via PCA to get a more dense representation that emphasizes "similar body shapes" with a blended gender perception of the dataset in space. - Similiarily, revisit points in the latent space, every time always visit the closest point to get a "morphing" of body shapes. Then (very) later visualize the gradual morphing/changes of body shapes via the corresponding 3d pose data.

(here we get only 1 morphing result)
Hopefully by comparing two ways of morphing results, one according to gender-specific attributes that are interpreted via human perception (linguistic) with underlying bias, another with merely non-gender-specific attributes with a general perception of how feminine/masculine the body shapes are with the help of PCA to see how such ways of exploring body type similarities could be raised without dual-gender perception in mind.36:[{"id":"b42504bb23bff4a3285c69158a25f0559021e72c78a100626e71968cfe2109571db4d484c77fba1b81adc21673ce61f3d9e833cc97e9f002faae9748b2d70937","href":"https://static.observableusercontent.com/files/b42504bb23bff4a3285c69158a25f0559021e72c78a100626e71968cfe2109571db4d484c77fba1b81adc21673ce61f3d9e833cc97e9f002faae9748b2d70937","name":"Screenshot 2022-12-12 at 3.26.32 PM.png","type":"image/png","lastModified":1670826456120.878,"size":155231},{"id":"c0e73562fcd7eaa2620fbe970444ad77d3b5235d2d70a9795b20c8a3de3499f886a47f44e26e8d0fe0e1bd475f6f592d327cadf2370d73278114373971b5e509","href":"https://static.observableusercontent.com/files/c0e73562fcd7eaa2620fbe970444ad77d3b5235d2d70a9795b20c8a3de3499f886a47f44e26e8d0fe0e1bd475f6f592d327cadf2370d73278114373971b5e509","name":"Screenshot 2022-12-12 at 3.26.02 PM.png","type":"image/png","lastModified":1670826456667.682,"size":768092},{"id":"871ada18da83ccd2a7f9c33e1b3f7a7aa46816e3315b7094f8c9138bb0ea09163132fe01c0f8f7708fa878163082228fcc40a0d8201eacf239265fe87c5e7634","href":"https://static.observableusercontent.com/files/871ada18da83ccd2a7f9c33e1b3f7a7aa46816e3315b7094f8c9138bb0ea09163132fe01c0f8f7708fa878163082228fcc40a0d8201eacf239265fe87c5e7634","name":"Screenshot 2022-12-12 at 3.26.13 PM.png","type":"image/png","lastModified":1670826456912.398,"size":1485905},{"id":"07c961f773cab5670659ab64b50f5c1232ff76ae48908d1d215d5e1900ed2455f08c67449b22fd4804fcac8ab2edf554c3789cb510603009df11bea529f8a24b","href":"https://static.observableusercontent.com/files/07c961f773cab5670659ab64b50f5c1232ff76ae48908d1d215d5e1900ed2455f08c67449b22fd4804fcac8ab2edf554c3789cb510603009df11bea529f8a24b","name":"Screenshot 2022-12-12 at 5.20.22 PM.png","type":"image/png","lastModified":1670833341196.013,"size":186860},{"id":"7417993181063081539ff998b3898d90b3e27b820d081728807f2d9eafbf0ebe9ff2e1e68a20395bfe4044b19afd61661349506c65b6ee8002f9680e6af1a817","href":"https://static.observableusercontent.com/files/7417993181063081539ff998b3898d90b3e27b820d081728807f2d9eafbf0ebe9ff2e1e68a20395bfe4044b19afd61661349506c65b6ee8002f9680e6af1a817","name":"Screenshot 2022-12-12 at 5.20.29 PM.png","type":"image/png","lastModified":1670833341179.41,"size":171900},{"id":"cd8604befdbc899c4335411f275e1767dead65d6b8dd515ad543a4a8371b345aa84d40b870fec6d4637c3578c6687f9c87c707c3c96132b3bd31e809360a84cb","href":"https://static.observableusercontent.com/files/cd8604befdbc899c4335411f275e1767dead65d6b8dd515ad543a4a8371b345aa84d40b870fec6d4637c3578c6687f9c87c707c3c96132b3bd31e809360a84cb","name":"Screenshot 2022-12-12 at 5.21.40 PM.png","type":"image/png","lastModified":1670833341382.434,"size":379307},{"id":"05c3bab908fe94de904bd1dfecb9dd9ceb002f87c686fe055e91ea49e44ef4baf9e7bae6fee4afcded2d970bb6a53f7ff83809062b94e260c02fecd59d9cd95a","href":"https://static.observableusercontent.com/files/05c3bab908fe94de904bd1dfecb9dd9ceb002f87c686fe055e91ea49e44ef4baf9e7bae6fee4afcded2d970bb6a53f7ff83809062b94e260c02fecd59d9cd95a","name":"Screenshot 2022-12-12 at 5.14.38 PM.png","type":"image/png","lastModified":1670833341455.764,"size":428343},{"id":"3ecb034298c06cd38562c32d1c91d24268d3a980b30f3b09a9b0a04187c2f831da515fa7e350938a5f83319d660a28c3176efc765db635ea747788b84a222ffc","href":"https://static.observableusercontent.com/files/3ecb034298c06cd38562c32d1c91d24268d3a980b30f3b09a9b0a04187c2f831da515fa7e350938a5f83319d660a28c3176efc765db635ea747788b84a222ffc","name":"Frame 6@1.png","type":"image/png","lastModified":1670836016370,"size":152970},{"id":"be495cd5e43b0a492c80ee18f01cbe67cf72d5f9661c7c7df9bca9d87c8f4738704e18bc4d8a23c856367ad0a5eab4c2b8b0542e204629ed42b0e8fb040dc097","href":"https://static.observableusercontent.com/files/be495cd5e43b0a492c80ee18f01cbe67cf72d5f9661c7c7df9bca9d87c8f4738704e18bc4d8a23c856367ad0a5eab4c2b8b0542e204629ed42b0e8fb040dc097","name":"gender specific.png","type":"image/png","lastModified":1670906302704.525,"size":136393},{"id":"17dedaa0fc5da8c0dcc2ca8bf65b385254e8b516000792d33d6d06c34235280228e11b6476aa6d5221180e3bb1ac05b87d2244f8352838b8e88afb1515b6cceb","href":"https://static.observableusercontent.com/files/17dedaa0fc5da8c0dcc2ca8bf65b385254e8b516000792d33d6d06c34235280228e11b6476aa6d5221180e3bb1ac05b87d2244f8352838b8e88afb1515b6cceb","name":"non-gender specific.png","type":"image/png","lastModified":1670906302521.22,"size":69034},{"id":"b51741431e4eadb038e474e674019bd7a57932ade4f525627d30df919b6b9c48da721ed01de00b4480b09942c786c4fb55d5d3bdd843f206454cc3b6a632308c","href":"https://static.observableusercontent.com/files/b51741431e4eadb038e474e674019bd7a57932ade4f525627d30df919b6b9c48da721ed01de00b4480b09942c786c4fb55d5d3bdd843f206454cc3b6a632308c","name":"Screenshot 2022-12-13 at 12.41.42 PM.png","type":"image/png","lastModified":1670902977088.755,"size":279232}] 31:{"id":"3f2542c33395a842","slug":"ml-for-artist-final-project","title":"ML for Artist Final Project","thumbnail":"40d38d424b17d01a45c2c82154a002f5c2b2160d4c400a4dc86a55cbc6aad709","access_level":"public","create_time":"$D2022-12-12T03:15:24.964Z","update_time":"$D2022-12-13T11:26:28.747Z","view_time":null,"first_public_time":"$D2022-12-12T03:15:24.964Z","has_unpublished_edits":null,"fork":false,"stars":1,"starred":false,"owner":{"type":"team","id":"c3d91d7db2867b4d","login":"betahsu","name":"ychsu6","avatar_url":"https://avatars-next.observableusercontent.com/197b7792c36f626b9fe391a472486b8b6dc3cf3503a2bb52e365e7916c324a9e","tier":"starter_2024"},"version":173,"publish_version":null,"role":null,"authored":false,"authors":[{"type":"individual","id":"8205ffe96d046f8c","login":"betahsu","name":"ychsu6","avatar_url":"https://avatars-next.observableusercontent.com/197b7792c36f626b9fe391a472486b8b6dc3cf3503a2bb52e365e7916c324a9e","tier":null}],"members":[{"type":"individual","id":"8205ffe96d046f8c","login":"betahsu","name":"ychsu6","avatar_url":"https://avatars-next.observableusercontent.com/197b7792c36f626b9fe391a472486b8b6dc3cf3503a2bb52e365e7916c324a9e","tier":null,"role":"editor"}],"comments":[],"custom_thumbnail":false,"trashed":false,"lapsed":null,"subscribed":null,"forks":1,"collection":null,"type":"notebook","body":{"title":"ML for Artist Final Project","cells":[{"id":0,"value":"# ML for Artist Final Project","pinned":false,"hidden":false,"mode":"md","author":"$undefined","summary":"$undefined","output":"","database":"$undefined"},{"id":5,"value":"## Topic I'm interested in\n- Gender Non-duality & Body Stereotypes","pinned":false,"hidden":false,"mode":"md","author":"$undefined","summary":"$undefined","output":"","database":"$undefined"},{"id":13,"value":"$33","pinned":false,"hidden":false,"mode":"md","author":"$undefined","summary":"$undefined","output":"","database":"$undefined"},{"id":57,"value":"$34","pinned":false,"hidden":false,"mode":"md","author":"$undefined","summary":"$undefined","output":"","database":"$undefined"},{"id":25,"value":"### Problems & Findings from this project \n- Anthropometric measurements of human bodies are surprisingly similar to the dark history of statistics introduced in class: **Measurements for the purpose of control** \n- The method of **Linguistic shape attributes** in fact, was inspired by [Body Talk](https://bodytalk.is.tue.mpg.de/overview.html), whose original intention, ironically, was to reveal **biases of linguistic descriptions of body shape** by visualizing body stereotypes with relating perceived 3D body shape. However, here the idea is used to solve real-world problems\n- The process of collecting such datasets is wrong in so many different ways yet the result seems to solve the problem so perfectly! It is working because the linguistic shape attributes perfectly align with human’s stereotypical perception of “body shape”. Therefore IT IS indeed working based on our idea of body shape, but does that mean it is right?","pinned":false,"hidden":false,"mode":"md","author":"$undefined","summary":"$undefined","output":"","database":"$undefined"},{"id":116,"value":"$35","pinned":false,"hidden":false,"mode":"md","author":"$undefined","summary":"$undefined","output":"","database":"$undefined"},{"id":22,"value":"## Artistic inspirations\n- **Jenny Saville**'s fat female nudes paintings question the \"positive\"/\"preferred\" female body image by showing the NOT-to-be-looked-at-ness bodies & forcing dual body characteristics in one \n- **Cindy Sherman**'s photography uses makeup and props to show how our gender perceptions are merely an acting, a disguise, a staged performance\n- **Judith Halberstam**'s the Queer Art of Failure\n- **Francesca Woodman**'s blurry figures photography work\n- **John Berger**’s Ways of Seeing\n- Movie: Portrait of a lady on fire ","pinned":false,"hidden":false,"mode":"md","author":"$undefined","summary":"$undefined","output":"","database":"$undefined"}],"files":"$W36","stdlib":"1","titling":"auto"}} 32:"$31:body"