First Principles of Line Drawings
This thesis presents an unsupervised method for creating line drawings from photographs or 3D models. Current methods often rely on high quality paired datasets to automate the creation of line drawings. We observe that line drawings are encodings of scene information that convey 3D shape and semant...
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Thesis |
| Published: |
Massachusetts Institute of Technology
2022
|
| Online Access: | https://hdl.handle.net/1721.1/139322 |
| _version_ | 1811097815108026368 |
|---|---|
| author | Chan, Caroline |
| author2 | Durand, Frédo |
| author_facet | Durand, Frédo Chan, Caroline |
| author_sort | Chan, Caroline |
| collection | MIT |
| description | This thesis presents an unsupervised method for creating line drawings from photographs or 3D models. Current methods often rely on high quality paired datasets to automate the creation of line drawings. We observe that line drawings are encodings of scene information that convey 3D shape and semantic meaning. We bake these observations into a set of first principle objectives and train an image translation network to map 3D objects into line drawings. We also explore generation of new styles of line drawings through a novel style confusion loss which averages and combines elements from different styles in a structured manner. User studies and quantitative experiments validate that our method encodes geometry and semantic information into line drawings and improves overall drawing quality. |
| first_indexed | 2024-09-23T17:05:23Z |
| format | Thesis |
| id | mit-1721.1/139322 |
| institution | Massachusetts Institute of Technology |
| last_indexed | 2024-09-23T17:05:23Z |
| publishDate | 2022 |
| publisher | Massachusetts Institute of Technology |
| record_format | dspace |
| spelling | mit-1721.1/1393222022-01-15T03:01:05Z First Principles of Line Drawings Chan, Caroline Durand, Frédo Massachusetts Institute of Technology. Department of Electrical Engineering and Computer Science This thesis presents an unsupervised method for creating line drawings from photographs or 3D models. Current methods often rely on high quality paired datasets to automate the creation of line drawings. We observe that line drawings are encodings of scene information that convey 3D shape and semantic meaning. We bake these observations into a set of first principle objectives and train an image translation network to map 3D objects into line drawings. We also explore generation of new styles of line drawings through a novel style confusion loss which averages and combines elements from different styles in a structured manner. User studies and quantitative experiments validate that our method encodes geometry and semantic information into line drawings and improves overall drawing quality. S.M. 2022-01-14T15:04:01Z 2022-01-14T15:04:01Z 2021-06 2021-06-24T19:18:10.999Z Thesis https://hdl.handle.net/1721.1/139322 In Copyright - Educational Use Permitted Copyright MIT http://rightsstatements.org/page/InC-EDU/1.0/ application/pdf Massachusetts Institute of Technology |
| spellingShingle | Chan, Caroline First Principles of Line Drawings |
| title | First Principles of Line Drawings |
| title_full | First Principles of Line Drawings |
| title_fullStr | First Principles of Line Drawings |
| title_full_unstemmed | First Principles of Line Drawings |
| title_short | First Principles of Line Drawings |
| title_sort | first principles of line drawings |
| url | https://hdl.handle.net/1721.1/139322 |
| work_keys_str_mv | AT chancaroline firstprinciplesoflinedrawings |