31:T173a,md`
# Faces of an Era ### by Yang Ren (2022)
[**webpage link**](https://yangren72.github.io/Faces-of-an-Era/website/)

#### Artist Statement:
**Overview**
Faces of an Era is a project that reflects on the identity, and history of the early Chinese immigrants that went to the USA in the period of gold rush. The project consists of multiple archive-looking looping videos that were generated by a machine learning algorithm. The constantly changing portrait photo of people and the "identity information" next to the photos are presented. Viewers are invited to browse those files generated by a software that was trained with archive files and portraits of the real people in the past. The artist intends to raise emotions and interests to history and the people through exploring the potential of software in the field of photographic imaging and history studies. **Process and Techniqes**
I first get the images from my [source](https://digitallibrary.californiahistoricalsociety.org/object/22481?solr_nav%5Bid%5D=74504d2e9f0c5c063c30&solr_nav%5Bpage%5D=2&solr_nav%5Boffset%5D=9#page/1/mode/2up). I input the images into lightroom to adjust the cropping, background, and resolution to prepare them for Ml training. I use StyleGan-2-ada with Pytorch to train my two models, one focuses on the portrait, the other focuses on the archive file. The instructions and the code, I take reference from [Jeff Heaton](https://github.com/jeffheaton) and [NVIDIA](https://github.com/NVlabs/stylegan2-ada-pytorch) I run the training on google colab. Later I explored the latent space and downloaded the generated image sequences. I then put the images into the after effect and combined the two videos(portraits and archive) together. I wrote a website with HTML and CSS to present the video loop. I trained my models on the Google Colab virtual machine. The code is a bit hard to show. So I decided to put two python files in Github that contain the code I use for Training and for Generation. **Concept**
Machine learning is a big part of the project in terms of techniques. However, it is seen by me as a creative tool to open more possibilities of image making, rather than problem solving. I read an article a few months ago talking about the Chinese immigrants in the early 1900s. I also found the database of the archive files. The history immediately interested me. After I learned about machine learning in the class, I came up with an idea of using the machine learning tool to create a generative program that can generate faces. The ML model directly learns from the real images and uses its "magic" to combine the details of people's faces together to create a new portrait. This process is fascinating when thinking about it in a photographic way. I see this as a special form of group portrait. It is like a photo collage, but arguably more random (less human interference) but also very smooth. When Interpolating the latent space, the transitions of the millions of faces and information inside it gives me a sense that it is bigger than life. It is bigger than individuals. It feels like it holds the spirit of the people in the past. It holds the spirit of history. The reason I choose to present it in a way of the archive file is that when it is presented with the archive files style image, the people are generalized with the format. The individuality is decreased and the sense of a group of people is strengthened. The viewers are distanced, which makes people think more about the history aspect of the story instead of a more personal one. I think it is more calm but profound. **Experiences**
The piece is present in a minimalistic website that has a simple interface of viewing. The archive style files are layout in the main page. All of the archive style files are moving slowly. The viewer can click on the small image to open up the image for a closer detailed look. The website is dark and clean, to replicate the feeling of browsing and trying to find the historical files. The interfaces are designed to allow the viewers to slowly look at whichever they prefer to look in whatever order they want. Each video loop is unique and the moving is quite slow forcing the viewer to calm down and look into the details just as what we do when we are looking into objective historical documents.

#### Screenshots

#### Github Source Code [Github page](https://github.com/yangren72/Faces-of-an-Era)

#### Video [video link](https://drive.google.com/file/d/1Ceno1towhfxUM04eE66zf4_A3eSTy0bR/view?usp=sharing)

#### Things to Improve on This is my first time trying to utilize machine learrning in my work. Thus, I encountered quite a lot of problem that is hard to solve with my current knowledge. Including using the trained model online for realtime generation and combining the two generated videos together with code. I also didn't have much experience in website coding. Therefore, it not only limits the way of how I can create my work but also how I can present it, which I think is an important part of this kind of work. I personally think this project is pretty challenging and experimental. In the process of training my model, I have already come up with some new ideas for machine learning related projects. I feel like this is a start of opening my knowledge to this catagory in which I would like to keep developing my skills and try out different things, including further developing this project(experiementing which more files from other source or using similar appraoch in other topics).

`32:T12b3,md`### 1. Project Idea:
**a. Change of old Project Idea:** I had do some research and some thinking about the final project proposal I planned to do and found quite a few concerns and flaws about my final project idea. One of them is that Chinese maps have long history and there are many variation of them, from small area to big, from artistic to highly technicle. I have fond that it is hard to choose my perspective to convey my idea of showing of the "Real River and Mountains from real map and at the same time retain an aesthetically pleasing reasult. If I show the real mountain, and river, it could look too much like an ordinary map, and it is hard to add other details other than mountains and river (traditional map have a lot of words on the ma to mark down the details). If I go on the other direction and create some details myself and try to make it not looking too much like a ordinary map, it is hard to implement the real river data into the map, and the map may sit on a strange position of half map half painting, which I don't prefer. Thus, I have been thinking of other alternative project idea that I want to do, and have came up with one idea after learning about machine learning last week. **b. New Project Idea:** Experiment with the photographic aspect of machine learning generated portraits. I want to generate portraits that represent the early immigrants of Chinese people that went to America for gold mining. Machine Learning in imaging is often used to create realistic fake photos. However, on the contrary, I am facinated by the idea that we can use machine learning to generate more abstract portraits with detailed retained, eventhough the datas feeded in are realistic photos. I like the idea of using machine learning to create abstract portraits rather than realistic images trying to replicate the original datas. August Sander is a photographer that documented the faces of people in all careers in the early 1900s. In his work, "Face of Our Time, he photographed people in way that is both aesthetically pleasing but at the same time reflecting on the society. The details of the wrincles on the face, how they wear and stand, and their emotions on the faces, all of those are captured in the collection of photos. It is an anthropology study.

In the 2000s century, photographer Ken Kitano having a similir starting point as Sander, created a photo series called "Our Face". However, he use the technique of overlapping photos of faces together, to create a somewhat "average" representation of the people. It aprears to be soulful and spiritual in my opinion.

I have always been inspired by Sander and Kitano and created portraits of poeple that replects the history or society. With the knowledge of machine learning technology, I have came up an idea to create portraits with programming that not only "represent" the people's faces I feed into the program, but also shows some aspect of expression that was created by the generative quality of machine learning. In the 19th century and early 20th century, many Chinese people from south east China, Hong Kong, Taiwan to California for gold mining. Many of these people went to America to seek for a better life. Many of them suffers from discriminations and unfair treatment at the time. They carried the race and culture to a new land. It is beautiful but sad [history](https://www.newyorker.com/culture/photo-booth/the-dark-purpose-behind-a-town-constables-journal). I found a [book](https://digitallibrary.californiahistoricalsociety.org/object/22481?solr_nav%5Bid%5D=74504d2e9f0c5c063c30&solr_nav%5Bpage%5D=2&solr_nav%5Boffset%5D=9#page/19/mode/2up) that uncommonly archived some ID style photos of these early immagrants. I want to use those photos to train a program that regenerate portraits what is somewhat abstract.

`33:T8ad,md` ### April 22nd - RISK AREAS
#### a. risk about GPU and programming langueages I want to train my own GAN model to be able to generate faces. however, I need GPUs to train for them. I have found that many people seems to use Google Colab for machine learning task since the free functions seems to be not bad. I am also planing to use google colab to work on my model. However, I have to learn some Python to be able to use it. I may face problem of that the google colab does not offer a good enough GPU for me to train. In this case, I may have to ask my friend to borrow his GPU or maybe I will have to borrow from Daniel. In case I cannot make a working Python program, I may also need to borrow GPU and build a program that work with Javascript.
#### b. risk of data collection The other thing that I have started to do and experiment is that the data from the [website](https://digitallibrary.californiahistoricalsociety.org/object/22481?solr_nav%5Bid%5D=74504d2e9f0c5c063c30&solr_nav%5Bpage%5D=2&solr_nav%5Boffset%5D=9#page/18/mode/2up) could be a bit hard to navigate by program because of the mouse navigation interface. Also, the position o the photo seems to be a bit inconsistent since the source was from old archive files. I may get inconsistent result from the scrapping program. I may consider screenshot the photo manually and resize them to fit perfectly in lightroom to get a consistent data for Machine learning training.
#### c. risk of the data not good enough/possible alternatives The other possible risk is that I may not have enough faces data to train it to create good result. In this case, I am thinking of other interesting "group portrait" source that I can still reflect on the photographic portrait aspect of machine learning. For example, photos the [people who got executed](https://www.tdcj.texas.gov/death_row/dr_executed_offenders.html) , people who joined HK protest, Taiwan Gay protest, or just a lot of photos I took of [people wearing mask on the Hong Kong Street](https://drive.google.com/file/d/1lUtTYBno2S0b6zp_VLB9y-9lQysYXZ3h/view?usp=sharing). when the virus just started. Those can create interesting results to experiment and explore.

`34:T1a09,md` ### April 22nd - Prototype
#### Prototype:
a. My current portotype is a trained [ML model](https://drive.google.com/file/d/1-iL-ZjHNqEd-N2bMmtqktVGyFgrYQMh2/view?usp=sharing) that can generate portraits of Chinese Miner. Also a group of still images.

b. The final version will have better presentation of the work with a better interface and also clearer delivery of the concept, hopefully a stronger implimentation of the concept and stronger energy and impact as well. Direction of improvement will be listed below.

#### Notes and the things I have done:
I have been training for my ML model for this week. I have found some very useful informations that helped me started my training. a course by [Jeff Heaton](https://sites.wustl.edu/jeffheaton/t81-558/) in washinton University helped me a lot. I have found that StyleGan2-ada-pytorch seems to be a good choice for training my model. According to their [Github page](https://github.com/NVlabs/stylegan2-ada-pytorch), it handles the overfitting problem better. ***"Training generative adversarial networks (GAN) using too little data typically leads to discriminator overfitting, causing training to diverge. We propose an adaptive discriminator augmentation mechanism that significantly stabilizes training in limited data regimes. The approach does not require changes to loss functions or network architectures, and is applicable both when training from scratch and when fine-tuning an existing GAN on another dataset."*** However, most people still tend to use a bigger dataset than mine. Thus, I have to pay more attention to maybe manually check if my model is getting towards an overfitting issue. Also, google colab is a bit hard to use. I have to train overnight and consistantly do snapshot to avoid disconnect problem. The generated portraits look much more realistic than I thought they could be, since I only have around 170 original photos as a dataset. Originally I actually imagined they would be more abstract and hoping for it. The realistic aspect of it kind of gives me a different opinion on the photos, whether it is good or not. Originally the faces look so similar and soulless and gradually they seem to be more and more soulful through the generations, which I think is a very interesting process.

ealier ones that looks similar

later ones that looks realistic

#### Things need to be done
the result differences than what I have thought. The photos are less abstract than I thought. I oringally thought that the abstract aspect of the faces could potentially create a sense of multiple identity combined together to represent a group portrait in a single image since it is obviously not real photos. However, the realistic aspect of it make it look like a real photo which really look like a real person. The higher dimension identity of combined soul doesn't seems to convey throught the realistic portraits. Thus, I have to think about what direction to go or a diffenret way to convey my original idea. Another thing I have noticed is that the process of forming the faces through the epochs are very interesting. As the photos I have shown above. The ealier photos looks somewhat empty, like a shell without much soul. But the later one looks soulful and less like a fake person. The process between the steps are interesting to me. I may also consider putting it in my work talking about the relation of soul and the images in photos. (it suddenly made me think of the core idea in blade runner. What makes a person real? What makes an human identity? How realistic should a person be to be seen as a real person?) Another interesting idea I have thought about, which Daniel also mentioned about in our conversation in email is to train the AI to generate the archive files instead of pure portraits. This to me is a very interesting but different approach to present the work. Seeing a portrait to me make people feel the emotion and energy more directly and strongly, people sees into the subjects' eyes and their expressions and through those details, we get a chance of seeing their background and identity. With archive files, the viewers are distanted by the format to sees the people who were somehow generalized and fit into the collection. We sees less of the emotion and have less subjectivity view into a person's appearance and expressions. Rather, we view them as a historical data that tells a bigger, less personal story of an era or incident. I think this is also a good way to present which makes the viewer thinks about very different things. If I choose to present it in a format of archive file, it is not actually generating a readable archive file. Although the data was from real people, but it is generating an image, or "painting" that gives the "feeling" of seeing a real person's archive data. It is interesting that we may sense the emotion of it as it is a "painting", but with a strong, superfical, hard-to-detach, more objective preconception of reading an archive file. This may result people to think about the historical background and value of the group of people in my hypothesis, which I think could be very interesting. Thus, I will keep training this new model and see how it goes.

A snapshot of an archive files style image that I am still training. As for how am I going to present my work in the final version, I think it would be different considering it as a portrait or as an archive file. In terms of the direction of making it as a "group portrait", I think with the more realistic style of generated results, I may need to expore the latent space and make it a video that shows the people's identity in a moving portrait, similar, but different from a traditional portrait photo. Or, maybe I will built a tweeter bot or website that can automatic generate a new portrait(still or moving) and put it in the website or on tweeter like a gallery or photo album. In terms of the direction of presenting it as an archive file, I may create an interface for people to view it online, or maybe also show the archive files as "painting" or "image" that present with a frame in an installation space or website? like how [Mario Klingemann](https://aiartists.org/mario-klingemann) present his "painting" in a form similar to paintings.

`35:T989,md` ### April 26nd - Some Progress #### Notes: I have been doing some more training on the archive syle files.

I have encountered some problem that the more I train, it does not necesserially gives me better result. I started to try figuring out what is going on. And find that sometimes my model's FID will go higher rather than the previous network and it may even break and create some nonsense.

I have done some [research](https://machinelearningmastery.com/how-to-implement-the-frechet-inception-distance-fid-from-scratch/) on the FID: ***The Frechet Inception Distance score, or FID for short, is a metric that calculates the distance between feature vectors calculated for real and generated images. The FID score is used to evaluate the quality of images generated by generative adversarial networks, and lower scores have been shown to correlate well with higher quality images.***

You can see the FID score goes from 87 to 271, then to even 333(lower is prefered in this case) after training a few more times . Also, the portraits in the archive file version does not seems to be as many variations as I wanted.

The portraits looks like similar persons, or should I say "siblings". I can think of some ways to try to experiment and potentially solve the problem. 1. Try to imporve my training. 2. train a model with archive files and another with only the portraits. Then, write a programe to sticks them together. 3. use selsected stills from different models that has different looks for the person's look. However, this case it will be rely more on manual work insdead of being abl to do more things automatically. For now, I am going to try to improve my training by giving a bigger dataset with data augumentation. I have found some codes online to flip, cracks, rotate the images, etc... However, since my data was quite small and very sensitive to rotation and flipping, I will try to create some contrast, color, optical distortions, grain, changes to the image with image processing software like lightroom, to create a bigger dataset and keep training on the existing model. `36:T5ee,md` ### May 5th some more progress
the data augmentation didn't work well. And the dealine is close, I have to decide on how to present the images. I then tried the latent vector interpolation and it looks quite nice, seems to be more abstract compared to the stills, which I think is a good balence that may make people thinks about both the historical contents and the emotions from the aesthetics. However, i may want it to change very slow. The fast change looks too showy, it doesn't got too much emotions. About presentation Originally I thought of putting the trained model online and generate in real time. I did a bit research and people recomended Flask. However, flask seems hard to learn as a python newbie. Thus, I then move on to creat a website that make it looks like an online archive library. Maybe i can create an interface that can view looping images. Also, since the further training didn't make the face looks more diverse in the archive model, I still want to use the face from the face model. I decided to use after effect to combined the images created by two models so that I have get the faces I want in a style of archive presentation. But in this way, I have to make the background white or colorful enough to be able to mask out eaily in after effect. I have to keep training the model with new database that has different color background for keying. At the end, I found the green one doesn't work well since the green poluted the images. Thus, I go with white at the end. `37:[{"id":"1bc4ae433af5f4ed9b17186de97bfbd1f5cb54e48441b6a0e24c19f7436a86e5e113d25c8e3e232979746cd8e888d89aa3823b6d45e0078a3e9e4ede8ac2ef66","href":"https://static.observableusercontent.com/files/1bc4ae433af5f4ed9b17186de97bfbd1f5cb54e48441b6a0e24c19f7436a86e5e113d25c8e3e232979746cd8e888d89aa3823b6d45e0078a3e9e4ede8ac2ef66","name":"website4.png","type":"image/png","lastModified":1652027001711.311,"size":1053312},{"id":"ea2c554adc1ebf4281a93cc99b82423665e8503323a28a142d53e8937058df4fe009132351fb9ba1ad9c7314e69afbe19b81c9b694de12bc017209e9ad42e845","href":"https://static.observableusercontent.com/files/ea2c554adc1ebf4281a93cc99b82423665e8503323a28a142d53e8937058df4fe009132351fb9ba1ad9c7314e69afbe19b81c9b694de12bc017209e9ad42e845","name":"website3.png","type":"image/png","lastModified":1652027002021.328,"size":990624},{"id":"6d79ec1d94bcc71b9e248d602306e39315c43a99db83f87320d630155242cb7911cbdc4f15b342b42758edd2ca37328f8d43574ab22d4e707497ff51050a2d9b","href":"https://static.observableusercontent.com/files/6d79ec1d94bcc71b9e248d602306e39315c43a99db83f87320d630155242cb7911cbdc4f15b342b42758edd2ca37328f8d43574ab22d4e707497ff51050a2d9b","name":"website1.png","type":"image/png","lastModified":1652027002370.144,"size":1625338},{"id":"99d259f5b8cb64ab165e6da83bc8d8f45c5c434fdb5c8b9045eb44b82f76b13e46de0551094e085373a4ebb6256f5841b8e50f0969291b5bb3caa9bcc08a672f","href":"https://static.observableusercontent.com/files/99d259f5b8cb64ab165e6da83bc8d8f45c5c434fdb5c8b9045eb44b82f76b13e46de0551094e085373a4ebb6256f5841b8e50f0969291b5bb3caa9bcc08a672f","name":"website2.png","type":"image/png","lastModified":1652027002362.816,"size":1079941},{"id":"ffe9ced218310786cd21041804cee3d0717ffab694a459236bdb14f00a6634f066faba4504adab9ddfa7bf2d1015e4abf8c5f6efacfaea5ad212259746a36da0","href":"https://static.observableusercontent.com/files/ffe9ced218310786cd21041804cee3d0717ffab694a459236bdb14f00a6634f066faba4504adab9ddfa7bf2d1015e4abf8c5f6efacfaea5ad212259746a36da0","name":"website5.png","type":"image/png","lastModified":1652044111246.688,"size":4306972},{"id":"aa1ccb8dfac17b9eecafa6869c9acc7aaca34a87f93b987352662d9f7c0b53d79a45eae697615bd0fa7e834d8dac76872a56c28c2a6cad67a61c5b421d0f42d7","href":"https://static.observableusercontent.com/files/aa1ccb8dfac17b9eecafa6869c9acc7aaca34a87f93b987352662d9f7c0b53d79a45eae697615bd0fa7e834d8dac76872a56c28c2a6cad67a61c5b421d0f42d7","name":"ch6.png","type":"image/png","lastModified":1649339596289.263,"size":2321068},{"id":"8efe86030a66cc3633d8d94d7d6faaa6449bf1ee1155eda3ac0c72f13c271e7eb5821e0b190f7bf270140093b3cab8cd26e172ee3d0de4c63df52e24256873e3","href":"https://static.observableusercontent.com/files/8efe86030a66cc3633d8d94d7d6faaa6449bf1ee1155eda3ac0c72f13c271e7eb5821e0b190f7bf270140093b3cab8cd26e172ee3d0de4c63df52e24256873e3","name":"august3.png","type":"image/png","lastModified":1649339594710.179,"size":563287},{"id":"288b8c1a97a97a8fc8b877fade99fddb13eb1de02f7ff5ad4715bb62360908f70366b96fd7554c2f97a9846dc1b3c4b9e18453bbabd378aa603c54c931c664db","href":"https://static.observableusercontent.com/files/288b8c1a97a97a8fc8b877fade99fddb13eb1de02f7ff5ad4715bb62360908f70366b96fd7554c2f97a9846dc1b3c4b9e18453bbabd378aa603c54c931c664db","name":"ken kitano3.png","type":"image/png","lastModified":1649339592273.856,"size":679121},{"id":"14e832794a6f18383ca82bfaa8baabe35c8402b6811d61e945d0246249cd15962942223917aea34fdf440dae1b6abc57d8762327755978e8630ae928ea64a150","href":"https://static.observableusercontent.com/files/14e832794a6f18383ca82bfaa8baabe35c8402b6811d61e945d0246249cd15962942223917aea34fdf440dae1b6abc57d8762327755978e8630ae928ea64a150","name":"ken kitano 2.png","type":"image/png","lastModified":1649339591723.735,"size":169606},{"id":"e8c27e96c16a15f8ed570978deaedda10570f88fc48491ef1ec72112e3a2787e8980bba02eab66aceacc466375538e0983075fe17dea0c42b69a24df60acb26e","href":"https://static.observableusercontent.com/files/e8c27e96c16a15f8ed570978deaedda10570f88fc48491ef1ec72112e3a2787e8980bba02eab66aceacc466375538e0983075fe17dea0c42b69a24df60acb26e","name":"ch5.png","type":"image/png","lastModified":1649339591494.759,"size":75651},{"id":"91bc560d0ac13b54aca52d8547a8422fef1b31fed0684304056e3f04b99d9f9d5bd828f27fa29b2a9b6fd9717df596240ed48def8a47d2fc2cce4d568b1e0828","href":"https://static.observableusercontent.com/files/91bc560d0ac13b54aca52d8547a8422fef1b31fed0684304056e3f04b99d9f9d5bd828f27fa29b2a9b6fd9717df596240ed48def8a47d2fc2cce4d568b1e0828","name":"ch2.png","type":"image/png","lastModified":1649339591322.125,"size":82291},{"id":"49f5bd9e3b848813eb846f3630537dddc4275c625e3c5b99bc69ec2c28ad58c9c1227b7ce3cb2bc06bea632e0d91b908d47ead07b7214a19e93be32c8f1af97b","href":"https://static.observableusercontent.com/files/49f5bd9e3b848813eb846f3630537dddc4275c625e3c5b99bc69ec2c28ad58c9c1227b7ce3cb2bc06bea632e0d91b908d47ead07b7214a19e93be32c8f1af97b","name":"ch4.png","type":"image/png","lastModified":1649339591162.103,"size":78149},{"id":"ed02428d3416bbca2b352739bf04fd70e031a575d69559050c1db4d925c4dd8875a47cbb14af512a7650f466fed30e1abcfb602744cf1f4ffeeb767f2ac861a0","href":"https://static.observableusercontent.com/files/ed02428d3416bbca2b352739bf04fd70e031a575d69559050c1db4d925c4dd8875a47cbb14af512a7650f466fed30e1abcfb602744cf1f4ffeeb767f2ac861a0","name":"august2.jpg","type":"image/jpeg","lastModified":1649339590961.038,"size":66939},{"id":"a5694ec39fca196c2ef45314d770d8337f12db124e3d7659d07b3c2ee9712b5d71aa10277a0d2ffc23d5f09876c30d671f1aff818060c63981fe2bcf55054505","href":"https://static.observableusercontent.com/files/a5694ec39fca196c2ef45314d770d8337f12db124e3d7659d07b3c2ee9712b5d71aa10277a0d2ffc23d5f09876c30d671f1aff818060c63981fe2bcf55054505","name":"ch3.png","type":"image/png","lastModified":1649339590953.953,"size":89919},{"id":"5369302fed4b4f7fc265f0f3b5d4cddd63f3b5e4b6da959b35c264175ca3131fcf7eb6d8106d968f8d2769425dd25e5592442708e5b19402876f0b87fff4d93d","href":"https://static.observableusercontent.com/files/5369302fed4b4f7fc265f0f3b5d4cddd63f3b5e4b6da959b35c264175ca3131fcf7eb6d8106d968f8d2769425dd25e5592442708e5b19402876f0b87fff4d93d","name":"ch.png","type":"image/png","lastModified":1649339590899.741,"size":87479},{"id":"1858d80366e6c0c8c2700e2b16947060f6ae481d27dfcaee4f366980f399061730d4f4dd4324d00ff35b8d0c87dcc055aa96a9e9d99d4b6736d503294d596160","href":"https://static.observableusercontent.com/files/1858d80366e6c0c8c2700e2b16947060f6ae481d27dfcaee4f366980f399061730d4f4dd4324d00ff35b8d0c87dcc055aa96a9e9d99d4b6736d503294d596160","name":"august.jpg","type":"image/jpeg","lastModified":1649339590812.401,"size":44520},{"id":"a23e90775bd40544031722fee19abff15f654a08e639b63681ff5c5ea4116232da649560f74892af594b6db970a7c1e6fb42035c85c9abc119ecc565a19ee3f8","href":"https://static.observableusercontent.com/files/a23e90775bd40544031722fee19abff15f654a08e639b63681ff5c5ea4116232da649560f74892af594b6db970a7c1e6fb42035c85c9abc119ecc565a19ee3f8","name":"ken kitano.jpg","type":"image/jpeg","lastModified":1649339590782.526,"size":84838},{"id":"b39aa9b71da74e07561a0df43456c198fa531c2377506855514ab527081ebbbaced9d12ac6c4cf35e8c44f0044ee9945bd04110e57c88593260271e48e0f648d","href":"https://static.observableusercontent.com/files/b39aa9b71da74e07561a0df43456c198fa531c2377506855514ab527081ebbbaced9d12ac6c4cf35e8c44f0044ee9945bd04110e57c88593260271e48e0f648d","name":"faces.png","type":"image/png","lastModified":1650636590856.952,"size":6934148},{"id":"f0d34d11787293fb205adca0afdd1942c680546d68b67b47d7df0d349507ada562f780a863b60168564bae2a91f410a814bca4ecede49bbf1968f08fbf3f2046","href":"https://static.observableusercontent.com/files/f0d34d11787293fb205adca0afdd1942c680546d68b67b47d7df0d349507ada562f780a863b60168564bae2a91f410a814bca4ecede49bbf1968f08fbf3f2046","name":"earlier.png","type":"image/png","lastModified":1650638616229.119,"size":4893887},{"id":"f888f7a28ed420720cdc6df653a72238c129d714ad2e666f010622d3a53e9816b0281cc1486ca9248ca553b4f05c4a80d72987e850677468a53e3fd1c4067d8a","href":"https://static.observableusercontent.com/files/f888f7a28ed420720cdc6df653a72238c129d714ad2e666f010622d3a53e9816b0281cc1486ca9248ca553b4f05c4a80d72987e850677468a53e3fd1c4067d8a","name":"later.png","type":"image/png","lastModified":1650638604105.883,"size":5126544},{"id":"7c2cc19439288a0c074c29ef3c75d8b395c4c29f3884b37de0207be69448c2a50919cdce5fb806c2440a6766408831f92933c9011d8a91a72b4009c1367f9948","href":"https://static.observableusercontent.com/files/7c2cc19439288a0c074c29ef3c75d8b395c4c29f3884b37de0207be69448c2a50919cdce5fb806c2440a6766408831f92933c9011d8a91a72b4009c1367f9948","name":"archive@1.png","type":"image/png","lastModified":1650641485267.625,"size":2333093},{"id":"40b559d463a5c565a289476c26f5b584888ea47618ff281ee876274e314085dede1abbc2e7d90c768647854de1b86b8bda083e87310444046a9cb5f468b049de","href":"https://static.observableusercontent.com/files/40b559d463a5c565a289476c26f5b584888ea47618ff281ee876274e314085dede1abbc2e7d90c768647854de1b86b8bda083e87310444046a9cb5f468b049de","name":"archive good.png","type":"image/png","lastModified":1650979134235.063,"size":2214274},{"id":"a9413df634709b65fcc5f2971b9d4aa2861b463c45c87840c379b634d4918f164b22630bd762cae4536de67750c0bb3e4376463b73d477bd7ad825f1ab85e95d","href":"https://static.observableusercontent.com/files/a9413df634709b65fcc5f2971b9d4aa2861b463c45c87840c379b634d4918f164b22630bd762cae4536de67750c0bb3e4376463b73d477bd7ad825f1ab85e95d","name":"archive good 1.png","type":"image/png","lastModified":1650979073248.515,"size":2229899},{"id":"0ecd0c2d8c76552dde166fb8ee15fe28bd044af594f5ec6199532d6fd2a49dfe5db211e60366be2794ef4342962ee9d273fa016f27b157779ddf3552cbaf7b31","href":"https://static.observableusercontent.com/files/0ecd0c2d8c76552dde166fb8ee15fe28bd044af594f5ec6199532d6fd2a49dfe5db211e60366be2794ef4342962ee9d273fa016f27b157779ddf3552cbaf7b31","name":"archive mess.png","type":"image/png","lastModified":1650979072393.589,"size":2146718},{"id":"08655541d0f27ccbcc15b88fe4f2a131d9aa4ef4ea467e9aef6407680bd703b6980d00e50a9098d6b5163aef8806bfc1fceb18261e59f7e9da4366d485f703a2","href":"https://static.observableusercontent.com/files/08655541d0f27ccbcc15b88fe4f2a131d9aa4ef4ea467e9aef6407680bd703b6980d00e50a9098d6b5163aef8806bfc1fceb18261e59f7e9da4366d485f703a2","name":"fid.png","type":"image/png","lastModified":1650979069630.891,"size":38784}] 2f:{"title":"Final Project","cells":[{"id":0,"value":"# Final Project","pinned":false,"hidden":false,"mode":"md"},{"id":126,"value":"$31","pinned":false,"hidden":false,"mode":"ojs"},{"id":6,"value":"$32","pinned":false,"hidden":false,"mode":"ojs"},{"id":37,"value":"md`### April 8th - FINAL PROJECT PECHA KUCHA:\n[Presentation Slide](https://docs.google.com/presentation/d/1rxUDOXK1aLHkb8ECetllY7R3xdPUWTrEVN3eVuWRro8/edit?usp=sharing)
\n[Presentation Video](https://drive.google.com/file/d/1Uwu5f-QqTpCTmDCENQDm82qwJ83OqjHp/view?usp=sharing)\n`","pinned":false,"hidden":false,"mode":"ojs"},{"id":32,"value":"$33","pinned":false,"hidden":false,"mode":"ojs"},{"id":48,"value":"$34","pinned":false,"hidden":false,"mode":"ojs"},{"id":100,"value":"$35","pinned":false,"hidden":false,"mode":"ojs"},{"id":121,"value":"$36","pinned":false,"hidden":false,"mode":"ojs"}],"files":"$W37","stdlib":"1","titling":"auto"}