metadata
license: apache-2.0
task_categories:
- mask-generation
- image-to-image
- image-to-text
tags:
- synthetic
- math
- graph
- latex
- function-graph
- image2latex
- function-plot
- mathematics
pretty_name: Elementary Functions MaskedGraph-LaTeX
configs:
- config_name: elementary_functions_onemask
data_files:
- split: train
path: ElementaryFunctions_OneMask/*.parquet
dataset_info:
features:
- name: graph
dtype: image
- name: masked_graph
dtype: image
- name: latex_formula
dtype: string
- config_name: complex_functions_multimask
data_files:
- split: train
path: ComplexFunctions_MultiMask/*.parquet
dataset_info:
features:
- name: graph
dtype: image
- name: masked_graph
dtype: image
- name: latex_formula
dtype: string
size_categories:
- 1K<n<10K
Elementary Functions MaskedGraph-LaTeX
This dataset was created to teach the model to "decompose" complex functions into elementary functions, isolating their pieces as elementary functions by associating masks with complex functions.
For elementary functions (such as cos, sin, x^n), a single mask was used. This way, the model understands that curvilinear and linear functions should only receive a single mask.
For more complex functions with multiple monomials, the graphs were highlighted with multiple masks of different colors, thus isolating the elementary functions.
In this way, the model learns to "decompose" complex functions into parts of simple functions.