Towards End-to-End Semantically Aware 2D Plot Tactile Generation
Bibliographic record
Abstract
An estimated 1.5 million Canadians identify as having vision loss, reaching up to 253 million people worldwide.The most effective way of delivering visual information to people with visual disabilities is through tactile graphics.Tactile graphics have textured or raised surface that makes them understandable by touching.To generate tactile graphics, a designer usually uses software such as CorelDraw to draw a grayscale image of the subject which will then be printed by an embosser.The goal of this thesis is to automate the process of converting 2D plots including Bézier curves, scatter plots, polygons, and bar charts to tactile format.Ultimately, our model can be used as an add-on to speed up the translation process that tactile designers execute.The end user of this model can be tactile designers not people with visual disability themselves, since implementing an end-to-end pipeline for tactile generation without the need to a designer is beyond the scope of this thesis.We defined the problem as an image-to-image translation task where the source domain belongs to 2D plots and the target domain is the tactile equivalent of the input plot.The proposed method is based on the pix2pix architecture which is one of the seminal works on paired image-to-image translation.The proposed models use UNet++ as the generator.We also propose to use gradient penalty and perceptual loss to further enhance the results.i To achieve editable outputs, we propose two approaches.One aims to generate RGB outputs.The other aims to generate multichannel outputs where each channel is associated with a component of the 2D plot.We evaluate the proposed models quantitatively and qualitatively.For RGB outputs we use foreground MSE, background MSE, precision, and recall.On the other hand, we use pixel accuracy, Dice coefficient, and Jaccard index to evaluate our channelwise model.On the combined category of Bézier curves, scatter plots, and polygons, the proposed channelwise method enhanced the pixel accuracy of the base model from 0.58 to 0.98.It also improved the Dice coefficient from 0.03 to 0.28, and Jaccard index from 0.01 to 0.21.Likewise, for the bar charts, the proposed channelwise method enhanced the pixel accuracy of the base model from 0.82 to 0.98.It enhanced the Dice coefficient from 0.06 to 0.31, and Jaccard index from 0.03 to 0.28.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.002 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.000 | 0.001 |
| Scholarly communication | 0.003 | 0.003 |
| Open science | 0.002 | 0.005 |
| Research integrity | 0.002 | 0.002 |
| Insufficient payload (model declined to judge) | 0.015 | 0.009 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".