Work Sample _ Hongfei Ji by Ji Hongfei

WORK SAMPLE OF HONGFEI JI

Master of Science in Computational Design School of Architecture College of Fine Arts Carnegie Mellon University hongfeij@andrew.cmu.edu

JI HONGFEI

+86 18817511675 | hongfeij@andrew.cmu.edu EDUCATION Tongji University Bachelor of Architecture • GPA: 4.77/ 5.0 National University of Singapore Non-graduating master program • GPA: 4.67/ 5.0 Carnegie Mellon University Master of Science in Computational Design

Shanghai, China 2016.09-2022.07 Singapore 2020.01–2020.05 Pittsburgh, the U.S. 2022.09–

Academic Experience & Internship Busan International Architecture Design Workshop & SUAE ASIA Busan, South Korea Participant | Culture Community Renovation Proposal, Under-bridge Space Research 2018.07–2018.08 Ruth & Arthur Scherbarth Art History Tour Bern, Switzerland & Berlin and Hamburg, Germany Participant |”The Sacred”, Presentation, Forum and Field Trip about religious artworks and relics 2019.08 META-PROJECT Beijing, China Assistant Architect | Landscape (full-charge) and architecture design of Xiaopu Culture Plaza 2020.06-2020.9 Archdaily CN Online Translator and Web Editor| Over 60000 words of translation from English to Chinese 2020.11-2021.11 Digital FUTURE 2021 Environmental AI in Design Group Shanghai, China Participant| Climate and Solar Energy-driven generative design proposal 2021.07 ARUP Integrated City Design Group Shanghai, China Assistant Urban Designer | TOD and Urban design, GIS data analysis, Art event advisory 2021.09-2021.12 Activities Tongji University Student Union Graphic Designer | Posters, magazines and print design Tongji University College of Architecture and Urban Planning Student Union Video Editing Department Leader | Video editing, special effect creating and trailer directing Award & Exhibition Tongji University Excellent Student Scholarship First Class 2 times/ Second Class 2 times National Scholarship

Shanghai, China 2016–2017 Shanghai, China 2017-2018

2017-2021

2018 Shanghai Urban Space Art Season Shanghai, China Exhibitor (on behalf of Tongji University Architecture Department) | Co-organized by UN-Habitat and the Shanghai Municipal Government 2017.10 Urban New Life - Steel Structure Creative Design Competition Shanghai, China Third Prize 2019.8 CAUP Best Studio Coursework Award & Exhibition of 2019 Shanghai, China Shortlist 2019.12 The NUS M.Arch Show 2020 Singapore Nominee 2020.06 SKILLS 2D&3D: Auto CAD, Adobe Suite (Photoshop, Illustrator, InDesign, Premier, Affect Effect), Rhino, Unity, Figma Programming: Python, QGIS, Grasshopper, JavaScript, Java Rendering: V-ray, Lumion, Enscape Languages: Fluent in English and German; Basic French; native Mandarin speaker Certificate: Machine Learning Foundations: A Case Study Approach by University of Washington - Coursera

CONTENTS

Pet Sensor

A sensor connecting human and pets, improving mental health and protecting privacy Hongkou District, Shanghai Individual Work B.Arch,Tongji University

Re-3D WuKang Road

An interactive AR framework to re-3D site history among fragmented web information Xuhui District, Shanghai Teamwork with Q.F.Yang, R.Shen, S.M.Chen Atom Atelier 2021 MR Workshop

Habitat 1000 - Rule-based Algorithm City A study on the metaverse entertainment space Virutal Space Teamwork with L.X.Zhao, Y.Q.Wang, Z.H.Xue, M.S.Wang, M.Z. Pan B.Arch Graduation Thesis,Tongji University

01 PET SENSOR A sensor connecting human's and pets' emotions, improving mental health under privacy safety

In the modern high-density metropolitan, the urban diseases tortured citizens despite the abundant physical life. In Shanghai, a part of the downtown is still a highdensity slum. More than two generations always live in a minimal house. They have to deal with daily conflict with family, safety, and sanitation problems. The living environment brings stress, but architects can do little about it. The Internet of Things is a modern technology collecting data to operate the appliance automatically. They can use advanced equipment to improve their living quality. However, for these people, who are suffering from the poor built environment, it's also urging to protect them from the harm of new biopolitics - the stealth of privacy by internet tycoons. Thus, the pet sensor is developed to collect human emotions indirectly and control the terminal equipment as their owners wish, which doesn't need an accurate operation but is agile without harming their privacy. Touched by the awful odor of slum because of the poor sanitation and humid climate in Shanghai, I chose the smell factor in a minimal house, where I finished my renovation design studio, as the example to test the proposal. But a similar approach can be applied to other forms of smart home applications.

Credits: 5 month, 2017.03-05 2021.09-12 (modified) Hongkou District, Shanghai Individual Work B.Arch,Tongji University Tutors: Kai Wang, Yanbo Li Tools: Figma, Apache Echarts, Rhino+Grasshopper, Arduino

PET SENSOR | Initiate - Why we need pet sensor?

More pet is adopted in cities with higher population density 3421.6 persons / sq km is the population density of Shanghai, 2020 0.1 pets / per person living in Shanghai, 2020 2

site:

Hongkou District, Shanghai

sample: a

Two-Generation Familiy with their Pet Dog

Pet Number per Capita Population Density

1. Biopolitics, Internet Tycoon and privacy - higher demand on human signal measurement Biopower is the "power that exerts a positive influence on life, that endeavors to administer, optimize, and multiply it, subjecting it to precise controls and comprehensive regulations." - Michel Foucault

In modern society, the executive body of Biopolitics has gradually transited from the government to Internet tycoons, who manipulate p e o p l e's l i ve s w i t h a m a s s i ve amount of data.

The EU General Data Protection Regulation (GDPR) went into effect on May 25, 2018. In recent years, the regulation is imitated by more and more countries. This policy also influences China, which puts up its version in 2021.

2. Agile and responsive - the limitation of architectural intervention and demand for new IoT sensor prototype The IoT devices are widely used on the appliance and improve our living environment beyond the architect's capability. However, the devices' functions are constrained by physical spaces and hard to control with the real-time human status change. Thus a new sensor prototype, agile and responsive, is demanded.

3. Pet interevention - when AI is still artificial

Artificial Intelligence hasn't stepped into every family, especially for the average living standard in developing countries. But a pet is affordable, adorable, and able to be the guard of its master.

PET SENSOR | OVERALL WORKFLOW OVERALL WORKFLOW

SIGNAL SELECTION

gesture

facial expression

odor

sound

visual From limited research papers, it's currently recognized that the most effective signal from me is the combination of gesture and sound. But for me, Visual signal is the most readable and accurate sign responding to human's emotion.

tactile

olfactory

acoustic

PET SENSOR | STEP 1: SIGNAL CAPTURE VOICE AND EMOTION - HUMAN Buddy, I feel so bored.

Stop! It's itchy!

Buddy, Let‘s run!

Zzzzz......

Eww...look at that!

I'm scared, Buddy.

DON'T DO THAT AGIAN! YOU HEAR ME?

......(Sobbing)

POSTURE AND EMOTION - DOG AXIAL

Ears: straight Angle: -30° — 30°

FORWAD

Ears: forward Angle: ≥30°

BACKWARD UP

Ears: backward Angle: -60° — -30°

BACKWARD DOWN

Ears: backward Angle: -90° — -60°

ASYMMETRIC LEFT

Ears: straight and backward Angle: -60° — 30°

ASYMMETRIC RIGHT

Ears: straight and forward Angle: -30° — 60°

OUT

WAGGING LEFT

MOVING SLOW

Tail: Out Direction and Angle : horizontal:-30° — 30° vertical:-30° — 30° Frequency: Breadth: -

Tail: Left Direction and Angle : horizontal: ≥30° vertical:-30° — 30° Frequency: Breadth: -

Tail: Out Direction and Angle : horizontal:-30° — 30° vertical:-30° — 30° Frequency: Low Breadth: Small

HIGH

WAGGING RIGHT

BROAD STROKES

Tail: Upward Direction and Angle : horizontal:-30° — 30° vertical:60° — 90° Frequency: Breadth: -

Tail: Right Direction and Angle : horizontal: ≤-30° vertical:-30° — 30° Frequency: Breadth: -

Tail: Out Direction and Angle : horizontal:-30° — 30° vertical:15° — 45° Frequency: Low Breadth: Large

DOWN

MOVING FAST

SHORT STROKES

Tail: Downward Direction and Angle : horizontal:-30° — 30° vertical:-60° — -90° Frequency: Breadth: -

Tail: Out Direction and Angle : horizontal:-30° — 30° vertical:-30° — 30° Frequency: High Breadth: Small

Tail: Out Direction and Angle : horizontal:-30° — 30° vertical:15° — 45° Frequency: High 03 Breadth: Medium

PET SENSOR | STEP 2 : DATA PARSING Reaction - Emotion - Aroma Dilution

When pets respond to their owner’s commands, the posture mentioned above will be recorded. The sensor only needs to record the performance (The voice of humans and the posture of dogs) and the operation that the device terminal should perform (diffuser). The cycle leverages the posture under the influence of their master as the clue, finds the corresponding emotion, and distills them to valence and arousal indicators to control the release of aroma. The intermediate process is finished previously by the cloud database, which simplifies the workflow to the greatest extent.

04 *

PET SENSOR| SENSOR DEVELOPMENT Based on the methodology of data collection and parsing. I proposed a sensor prototype, which is minimal and flexible. The units can be attached to any wearable accessories with the change of season and the difference between dog breeds (those have a deficiency on their ear(e.g., Labrador retriever) or tail(e.g., Corgi) characteristics).

1 Main Control and Transmission System

Dupont Line

Bread Board

ESP 32

2 Measurement System - Voice

3 Measurement System - Posture

1 Signal captured by ESP32 bluetooth module

2 Measurement system - Posture

4 Main control and transmission system

3 Measurement system - Voice

Feedback and Toggle ON The feedback helps both people and pets be emotionally stable

Measurement System The MPU 6050 can captures pet's tail and ear motion, Audio Anaylsis captures human's voice.

Audio Analyzer Module Audio Signal Capture

Analog Sound Sensor

MPU 6050 Module

Posture Detection

MPU 6050

Analog Sound Sensor

Audio Analyser

Main Control and Transmission System

The system turns on when pets approach their master by receiving wifi signal, and the collected data is sent to the cloud server.

ESP 32

Data Parsing

For Terminal Equipment Data processing is conducted based on the classification above to adapt to machine learning database.

Voice Raw Data

Posture Raw Data

Emotion

PET SENSOR| PROTOTYPE AND DEMO TEST Prototype Design

The sensor is flexible and can be pinned on any fabric gear for pets.

The sensor is integrated into bands, which is easy to tie on pets' ears or tails.

Demo Component

Component Assemble

Programming

Visualization

Test with a dog

Inteface

05 *

PET SENSOR |STEP 3: TERMINAL EQUIPMENT - Meditation App Interfa

OPERATION

MEDITATION

Users can set their meditation time manually here

Users can change meditation mode, background music or quit the meditation session

EMOTION SCOR

The emotion score generated based on th indirectly collected by

ace

e will he data y sensor

PET HEALTH SCORE

AROMA FORMULA

The pet health score will generated based on the data collected directly by the sensor

Current diffused aroma formular will be generated with the health status and advice, users can print it or share it on social media

06 *

02 Re-3D WuKang Road An interactive AR framework to re-3D site history among fragmented web information

WuKang Road is famous for its history as a french concession in 20th century Shanghai, the earliest metropolitan in China. Many tourists come here to feel the exotic atmosphere. Thus it brings the so-called 'influencer economy' to the street. However, the spread of internet platforms like Red (A Chinese social media platform) makes people believe that they can explore the site without being there. It's partly true, but the other part of value locates in the numerous historical scenes, architecture, and famous people. The homogenization of scenic spots caused by the Internet makes people almost forget that. Compared with the power of the Internet, the current boring guidance has no attraction to the tourist. Th u s , o u r go a l i s s i m p l e . S o m e m et h o d s s h o u l d b e implemented to regain people's interest in the real Wukang Road. We choose the AR app as our tool, combining with treasure hunt game and interactive AR filter. The objects from this site can be collected and used for further exhibition creation. Tourists can share and enjoy their art pieces and explore every corner of the street through AR guide. The app will also be a product of the collaboration of government, shop owners, artists, and tourists to make it a whole platform for all to learn what happened here.

Credits: 3 month, 2021.8-10 Xuhui District, Shanghai Teamwork with Q.F.Yang, R.Shen, S.M.Chen Role in Team: Conceptual Design(75%) Interface Design and Demo making (Individual) Atom Atelier 2021 MR Workshop Tutors: Xu Zhang, Senna, Yichen Jia Tools: QGIS, Figma, Unity3D, Aigsoft Metashape

The full UI interface prototype: https://www.figma.com/proto/esfTcCJyVtSMbiLss8L80a/ WK-Road?node-id=126%3A4846&scaling=scale-down&pageid=0%3A1&starting-point-node-id=126%3A4846&show-proto-sidebar=1

RE-3D WK ROAD | Problem Statement - Unmemorized and diluted Histo Observed from history, time, and popularity mapping, the problems can be summarized as 3 points below: 1. The DAKA disperses the real value of a historical site. The appearance of buildings is excessively emphasized. 3D space experience is overwhelmed by 2D. 2. Tourists' attention is drawn by several core visiting points, others are ignored. 3. The influencer moments iterate quickly in the last decade. No memories are recorded. This means no new history is created and remembered.

ory

08 *

RE-3D WK ROAD | Proposal and Design Process To solve the problem above, a process based on an AR miniapp is developed. It combines collection, curation, and share. The users can enjoy the treasure hunt game, curate their exhibition and interact with the tourists nearby. Internet entertainment is implemented on-site. Visitors focus on the real scenery points again and begin to learn the knowledge, history, and culture behind the delicate appearance of landmark architecture and c a re f u l l y re t o u c h e d p h o t o s . Besides, they can also realize their dream of creating an attractive selfie to share with their friends on social media.

The Brand and Shop Owner The brand and shop owner need to increase their popularity to assure their survival. Thus it’s a good deal to leave popular fragments to let tourists find the value behind.

The Government

The Artist

The government, especially the culture and tourist department, needs a new system to attract the tourists and takes the responsibility to let tourists learn the history behind Wukang Road.

The artists need a famous site to let their artwork known by more people. Thus they are willing to add something original to the new system

The tourists are hardly attracted by current boring introduction, which is also hard to find and read. They need a new method to learn the scenery point in a joyful way.

The Tourist

Historical Asset

Influencer Fragments

Tourists' Curation

Artworks

Historical Template

Influencer Template

Undefined Template

Art Exhibition Template

Technical Support

Skateholders' need

Tourists' Experience

Artists' advice

09 *

RE-3D WK ROAD | UI Interface The UI Interface is developed to realize the functions mentioned above - collect, curate and share. Site map as the core of the app, an AR navigation lead users to find asset (objects and exhibitions) and re-create with exhibitions curated by other on-site users. Users can leverage the asset they have found to take photos and share them on social media.

Select template

Make photos

Switch to Selfie Mode

Camera

Users can choose their favorite collection as a unique filter template, then they can take photos as a memorial or share with their friends as DAKA procedure. When they want to become creators of memory and history, they can also curate and upload their templates as exhibitions for other tourists to find out and re-create.

Camera - Share Collection

Check detailed information

Map

Check detailed information

Collect all asset

Redeem souvenir

Users can find their "trophies" (objects and exhibitions) here. When one of those is selected, the detailed introduction will appear to make the users learn the story behind it. When all official assets are collected, a redeem code will be generated. Then the user can receive an AR-interactive souvenir at the tourist center as a reward.

Collection

Check detailed information

User Upload Photogrammetry Tutorial

Grasp to collect

Find collection Find collection

Stretch to browse exhibitions

Tap to selected exhibiton

Re-create with selected exhibition

Treasure Hunt game to collect unfound assets

Make photos with selected exhibition

With the navigation, users can locate the assets and exhibitions and collect them under AR mode. They can also upload what interests them with the app's guidance. Each core point limits an area where users can interact with others on-site, re-create others' exhibitions or take photos with them. When they discover what they don't have yet, a treasure hunt game (AR navigation) will lead users to the destination. Thus they can explore more and learn more about the history and stories here.

Map - Curation 10

RE-3D WK ROAD | Technical Workflow and Demo Mockup Under the augmented reality technology, 3 ancillary skills are used here to realize the anticipated functions: 1. Photogrammetry: All the objects are scanned and modeled by Agisoft Metashape, then uploaded for later use. 2. Markerless AR Collection: The valuable historical sites and influencer stores are located manually. The objects are put there by SDK C# script. Posture recognition is implemented to ensure the AR gesture collection. 3. AR Navigation: The whole street is scanned and modeled as point clouds, the location of AR objects is used to generate the path. A series of UI interfaces and visualization lead users to their destination.

PHOTOGRAMMETRY Photos or Video Recording

MARKERLESS AR COLLECTION Script programming and model making

Objects Localisation

SDK code relates location with object's icon

Marker selcetion based on historical sites and popular POI

ARKit SDK Motion Capture

Photo selection & processing AR NAVIGATION Mapping

Model Generation by Agisoft Metashape

Scanning environment

Creating points clouds

Localisation

Data Model

Finding positions in point clouds

Wayfinding

Posture Recognition

Template fit user's status by body tracking

Visualisation

Trigger interaction by gesture

AR Objects in the previous locations

AR Guide Test

Visualisation

UI Interface

Visaulizing path

Occlusion

AR Collection Test

11 *

RE-3D WK ROAD | History creators - an interactive platform completed Crowdsourcing Upload

1. AR Objects Collection

2. Curate and user upload

d and improved by all

3. Interactive Re-creation

4. AR Navigation

12 *

03 Habitat 1000 - Rule-based Algorithm City A study on the metaverse entertainment space system

As the Covid-19 pandemic spread across the world, the physical connection between people was disrupted. Thus, the concept of "metaverse" based on virtual space began to thrive, and virtual reality technology, which had been developed for half a century, began to receive attention, and architects also took advantage of this trend, putting the computational design methods they had studied for dozens of years into practice. This project is based on the Shanghai's Covid-19 outbreak. Through observing people's mental needs during this period, this project translates the theme of the project - Habitat 1000 - into a spiritual dwelling, and proposes a new metaverse space system with reference to existing architectural algorithms, generation rules and interaction mechanisms. And taking the two potential functional demands of people for virtual space - outdoor performance and natural production experience, the design methodology of the form, mechanism and tools of virtual transportation, public space and complex function space under the new system is explored, which provides a new approach for the design of the metaverse space.

Credits: 4 month, 2022.3-6 Teamwork with L.X.Zhao, Y.Q.Wang, Z.H.Xue, M.S.Wang, M.Z. Pan Role in Team: Conceptual Design(75%) Coding (100%) System Design (90%) Unit Design(Individual) Graduation Thesis Tutors: Xiang Wang Tools: Python, Grasshopper, Rhino, Adobe Suite

Habitat 1000 | Workflow Diagram Space System under Metaverse

1. Traditional building system: Available: Space Hierarchy Not Available: Structure System based on Physics Laws

2. Internet(Web 2.0) Features Available: Keyword "Label" Unavailable: 2D Data Transmission Network

3. Organize "Labels" through spatial hierarchy, establish connections based on functional similarity, connect users with similar potential functional needs as social organization.

1. Word2Vec word cloud, determine the relative position of clusters.

2. The common Labels of clusters build a traffic core and establish a spatial hierarchy.

3. The relative position remains unchanged, but the whole system changes dynamically

Space System Generation Diagram

1. Social Media Comment Scraping

2. Comment keyword (Label) capture

3. Text Preprocessing Remove irrelevant/ interfering words

4. Keyword vectorization Python - gensim library: word embedding & word2vec algorithm

5. Tensorboard Embedding Projector vector visualization

6. Coordinate remap by rhino and grasshopper

Space System Generation Rules and Algorithm Diagram

Cluster

Low-speed Sightseeing Route

Public Space

High-speed Pipe Route

Traffic Core & Entry Point

Teleportation

1. Set the total number and density of voxel units, and generate a point cloud as the spatial domain for subsequent operations.

2. Restrict the boundary of the cluster keywords and traffic node keywords, limit the cluster keywords to the periphery , traffic core keywords to the kernel.

Surface Envelope Distribution + Quick Connection

3. Choose points and sort them according to the frequency. The keywords with high frequency are selected as the center of the cluster /public space.

4. Use kangaroo to avoid intersection, and project keywords to the point of the corresponding nearest spatial voxel.

4 6

5. Connect the cluster center and t h e t raf f i c co re a s t h e l ow - s p e e d transportation system.

6. Connect shortest path between all keywords as the high-speed transportation system and project to the previous coordinate system.

7. According to the location of the center of the cluster/public space and the unit number, find the nearest point as the cluster boundary.

8. Use the KNN algorithm to cluster with the current 4 public space clusters, and assign the remaining public space keyword to them.

Habitat 1000 |Word2Vec Algorithm Application and Visualization

Habitat 1000 | Prototype Generation and Iteration

Prototype 01

Prototype 02

Unit Density：2% Transit system proportion：40% Public node distribution：75% Unit number seed 10 Keyword combination 01

Unit Density：3% Transit system proportion：40% Public node distribution：75% Unit number seed 10 Keyword combination 01

U Tr P U K

Prototype 06

Prototype 07

Unit Density：4% Transit system proportion：20% Public node distributio：75% Unit number seed 10 Keyword combination 01

Unit Density：4% Transit system proportion：40% Public node distribution：75% Unit number seed 10 Keyword combination 01

U Tr P U K

Prototype 11

Prototype 12

Unit Density：4% Transit system proportion：40% Public node distribution：30% Unit number seed 10 Keyword combination 01

Unit Density：4% Transit system proportion：40% Public node distribution：45% Unit number seed 10 Keyword combination 01

U Tr P U K

Prototype 16

Prototype 17

Unit Density：4% Transit system proportion：40% Public node distribution：75% Unit number seed 10 Keyword combination 01

Unit Density：4% Transit system proportion：40% Public node distribution：75% Unit number seed 20 Keyword combination 01

U Tr P U K

Prototype 21

Prototype 22

Unit Density：4% Transit system proportion：40% Public node distribution：75% Unit number seed 10 Keyword combination 01

Unit Density：4% Transit system proportion：40% Public node distribution：75% Unit number seed 10 Keyword combination 02

U Tr P U K

Prototype 03

Prototype 04

Prototype 05

Unit Density：4% Transit system proportion：40% Public node distribution：75% Unit number seed 10 Keyword combination 01

Unit Density：5% Transit system proportion：40% Public node distribution：75% Unit number seed 10 Keyword combination 01

Unit Density：10% Transit system proportion：40% Public node distribution：75% Unit number seed 10 Keyword combination 01

Prototype 08

Prototype 09

Prototype 10

Unit Density：4% Transit system proportion：60% Public node distribution：75% Unit number seed 10 Keyword combination 01

Unit Density：4% Transit system proportion：80% Public node distribution：75% Unit number seed 10 Keyword combination 01

Unit Density：4% Transit system proportion：10 0% Public node distribution：75% Unit number seed 10 Keyword combination 01

Prototype 13

Prototype 14

Prototype 15

Unit Density：4% Transit system proportion：40% Public node distribution：60% Unit number seed 10 Keyword combination 01

Unit Density：4% Transit system proportion：40% Public node distribution：75% Unit number seed 10 Keyword combination 01

Unit Density：4% Transit system proportion：40% Public node distribution：90% Unit number seed 10 Keyword combination 01

Prototype 18

Prototype 19

Prototype 20

Unit Density：4% Transit system proportion：40% Public node distribution：75% Unit number seed 30 Keyword combination 01

Unit Density：4% Transit system proportion：40% Public node distribution：75% Unit number seed 40 Keyword combination 01

Unit Density：4% Transit system proportion：40% Public node distribution：75% Unit number seed 50 Keyword combination 01

Prototype 23

Prototype 24

Prototype 25

Unit Density：4% Transit system proportion：40% Public node distribution：75% Unit number seed 10 Keyword combination 03

Unit Density：4% Transit system proportion：40% Public node distribution：75% Unit number seed 10 Keyword combination 04

Unit Density：4% Transit system proportion：40% Public node distribution：75% Unit number seed 10 Keyword combination 05