Generate Images from Lines Lub Cuab Yeej Tsim Qauv Kev Txawj Ntse Sab Hauv Tsev los ntawm Junming Chen

Qhov qauv tsim Generate Images from Lines Lub Cuab Yeej Tsim Qauv Kev Txawj Ntse Sab Hauv Tsev los ntawm Junming Chen yog tus yeej muaj nqis heev ntawm qhov khoom plig 2023 A' Design Award Hlau hauv Pawg Sib Tw Tsim Kev Tsim Tawm, Kev Siv Algorithm, Parametric thiab AI Pab Tsim.

HlauKev Tsim Tawm· 2023

Generate Images from Lines Lub Cuab Yeej Tsim Qauv Kev Txawj Ntse Sab Hauv Tsev

los ntawm Junming Chen

Tus Neeg Siv: Junming Chen

Tus Yeej A' Design Award Hlau hauv Pawg Kev Tsim Tawm

47,588

Tus Yeej Khoom Plig Tau Txheeb Xyuas

Txheeb Xyuas Qhov Xwm Txheej Khoom Plig rau #150690

Saib ntawm A' Design Award Tsab Xov Xwm Daim Phiaj Khoom Plig

Xim

Qhov Tsim

Txog Generate Images from Lines Lub Cuab Yeej Tsim Qauv Kev Txawj Ntse Sab Hauv Tsev

Tsim Duab Los Ntawm Kab Kos yog ib lub kaw lus tsim tswv yim (artificial intelligence) uas tsim tau cov duab qauv kev txawj ntse sab hauv tsev los ntawm cov kab kos duab. Lub kaw lus no yog tsim los ntawm ib qho qauv "diffusion" uas tau txhim kho tshiab, thiab tau qhia los ntawm ntau tshaj 20,000 daim duab qauv kev txawj ntse sab hauv tsev uas cov kws tsim qauv muaj kev paub ntsis tau sau sib sau thiab teev cim tseg. Cov kws tsim qauv muab ib daim kab kos uas kos tes lossis kos hauv computer, thiab xaiv ib qho qauv zoo li niaj hnub no, Nyij Pooj, Tebchaws Europe, lossis Suav teb. Ib daim duab qauv yuav tsim tau tsuas yog ob peb feeb xwb. Kev tsim qauv hloov mus los ntawm kev kho cov kab kos qub, yog li tsis tas yuav siv tswv yim los tsim qauv thiab tsim duab qauv dua ib zaug ntxiv.

Qhov Lees Paub

Txog A' Design Award Hlau

Generate Images from Lines Lub Cuab Yeej Tsim Qauv Kev Txawj Ntse Sab Hauv Tsev yog tus yeej ntawm A' Design Award Hlau, ib qho kev lees paub muaj meej mom uas muab rau cov qauv tsim zoo uas ua tau raws li cov kev xav tau ntawm kev ua haujlwm thiab kev lag luam uas xav tau los ntawm cov khoom tsim zoo. Cov neeg yeej Hlau qhia txog kev txawj loj, kev paub tshwj xeeb siab, kev nkag siab ntawm cov kws tshaj lij thiab lub peev xwm tsim ntawm cov neeg tsim. Generate Images from Lines Lub Cuab Yeej Tsim Qauv Kev Txawj Ntse Sab Hauv Tsev xav tias yuav yog ib qho khoom tsim uas neeg hwm, qhuas thiab hlub uas siv tau thiab tsim tshiab. Cov qauv tsim yeej A' Design Award Hlau no sib xyaw cov kev coj ua zoo tshaj plaws ntawm kev lag luam thiab muaj cov yam ntxwv siv tshuab txawj ntse txuas nrog kev paub thiab txuj ci. Generate Images from Lines Lub Cuab Yeej Tsim Qauv Kev Txawj Ntse Sab Hauv Tsev tau txais kev hwm vim nws tag nrho thiab tsim los muab kev txhim kho rau lub neej uas pab ua rau lub ntiaj teb zoo dua.

Tom Qab Qhov Tsim

Pob Xov Xwm

Cov credit, cov neeg pab thiab cov ntsiab lus kev txawj ntse muab los ntawm tus tsim.

Yam Tshwj Xeeb: The design was done by artificial intelligence (AI) in collaboration with designers. We trained an AI that professionally generates interior designs, which can generate interior designs with different decorative styles based on input line drafts. It solves the problem that the traditional design modification requires remodeling and rendering. It only takes a few seconds to generate a design using this method, and the efficiency of generation and modification is increased by dozens of times.
Qhov Loj & Ntim: Use artificial intelligence to cooperate with designers to generate designs with different decorative styles. The design area is 160 square meters, and the decoration styles include modern, Japanese, European, and Chinese styles.
Cov Ntsiab Lus: Artificial intelligence, Human-machine collaboration, Interior design, Generative design, Efficient design, Rapid modification, Controllable design, Decoration style control
Thev Naus Laus Zis Tsim Khoom: The AI technology we use is our improved diffusion model technology. The typical defect of the diffusion model is that it cannot generate design according to specific "decorative style" and "space attribute," which is fatal for interior design. To this end, we re-collected and manufactured a data set containing more than 20,000 interior design renderings and data annotations and handed the data set to AI learning to generate an AI suitable for interior design. The AI can generate interior renderings specifying "decorative style" and "space attributes" by inputting line sketches.
Kev Sib Tw Tsim Qauv: The challenge of this design is to build a professional interior design AI that generates designs of high quality and correct decor. To this end, we asked professional designers to collect and label tens of thousands of interior design images and then handed over the filtered data to AI for training. Make AI understand what is "decorative style," "spatial attributes," and "aesthetic attributes" and generate an interior design with a specified decorative style.
Kev Tshawb Fawb: Research Background: The interior design process could be more convenient, and the efficiency of design and modification could be higher.
Methods: We trained an AI for professional interior design and used AI to cooperate with people to generate interior designs of various decorative styles from line drawings.
Results: Using our AI, designers can obtain an interior design rendering every few seconds, efficiently generating and modifying designs.
Insights: Our method can quickly generate large batches of interior designs with specified decoration styles and spatial attributes, and the generated results can be changed by modifying the line draft. Design efficiency is improved by eliminating the need for manual modeling and rendering.
Kev Txhawb Siab: The traditional process of obtaining renderings for interior design is tedious and time-consuming. We use artificial intelligence (AI) to generate interior design, improve design efficiency, and explore new ways of human-computer collaborative design. Specifically, we trained an AI to create interior designs, which can quickly generate and modify designs in cooperation with humans. Changing the design does not require modifying the model and re-rendering, which improves design efficiency.
Lub Sij Hawm Ua Hauj Lwm: The project is located in Guangzhou, Guangdong Province, China, and was designed in 2023.
Txheej Txheem Ua Hauj Lwm: 1. Professional designers collect interior design images and mark and screen them.
2. Use the filtered data to train a professional interior design AI.
3. Designers use AI to design.
4. The designer inputs the hand-drawn or design line draft into AI to specify the decoration style.
5. AI generates the interior design of the specified decoration style.
6. The designer adjusts the line draft of the unsatisfactory part of the design and then regenerates the design.
7. After adjusting the line draft and generating it multiple times, a good design is obtained.