FigMirror Data-Preserving Code Augmentation · 10 Cases

Each case compares two code-generated images. The right-hand figure is not a new task: it keeps the original data, labels, chart type/topology, and analytical relation, then applies FigMirror presentation improvements.

heatmap · Chart2Code_level2_contour_5_v5

Original code render versus data-preserving FigMirror code augmentation

Both sides are generated from code with a fixed random seed where needed. Left: unchanged original.py. Right: augmented.py, preserving the same data and chart topology while improving figure aesthetics.

original code renderoriginal codedataset/source imageaugmented code renderaugmented codecomparison imageprocess note

network · Chart2Code_level2_combination_3_v1

Original code render versus data-preserving FigMirror code augmentation

Both sides are generated from code with a fixed random seed where needed. Left: unchanged original.py. Right: augmented.py, preserving the same data and chart topology while improving figure aesthetics.

original code renderoriginal codedataset/source imageaugmented code renderaugmented codecomparison imageprocess note

surface3d · ChartNet-sample_2394253dc8502926

Original code render versus data-preserving FigMirror code augmentation

Both sides are generated from code with a fixed random seed where needed. Left: unchanged original.py. Right: augmented.py, preserving the same data and chart topology while improving figure aesthetics.

original code renderoriginal codedataset/source imageaugmented code renderaugmented codecomparison imageprocess note

box · Chart2NCode_0e7b573addd14f83aad26cf57f08876f

Original code render versus data-preserving FigMirror code augmentation

Both sides are generated from code with a fixed random seed where needed. Left: unchanged original.py. Right: augmented.py, preserving the same data and chart topology while improving figure aesthetics.

original code renderoriginal codedataset/source imageaugmented code renderaugmented codecomparison imageprocess note

contour · Chart2Code_level2_contour_25_v2

Original code render versus data-preserving FigMirror code augmentation

Both sides are generated from code with a fixed random seed where needed. Left: unchanged original.py. Right: augmented.py, preserving the same data and chart topology while improving figure aesthetics.

original code renderoriginal codedataset/source imageaugmented code renderaugmented codecomparison imageprocess note

radar · Chart2NCode_c372e9e299c44a6fa0506a700b87f9db

Original code render versus data-preserving FigMirror code augmentation

Both sides are generated from code with a fixed random seed where needed. Left: unchanged original.py. Right: augmented.py, preserving the same data and chart topology while improving figure aesthetics.

original code renderoriginal codedataset/source imageaugmented code renderaugmented codecomparison imageprocess note

violin · Chart2NCode_824d88ec85314c6ea8378d9e2e62e046

Original code render versus data-preserving FigMirror code augmentation

Both sides are generated from code with a fixed random seed where needed. Left: unchanged original.py. Right: augmented.py, preserving the same data and chart topology while improving figure aesthetics.

original code renderoriginal codedataset/source imageaugmented code renderaugmented codecomparison imageprocess note

tilemap · ChartNet-sample_15e07f5f51b5f852

Original code render versus data-preserving FigMirror code augmentation

Both sides are generated from code with a fixed random seed where needed. Left: unchanged original.py. Right: augmented.py, preserving the same data and chart topology while improving figure aesthetics.

original code renderoriginal codedataset/source imageaugmented code renderaugmented codecomparison imageprocess note

area · Chart2Code_level2_area_4_v5

Original code render versus data-preserving FigMirror code augmentation

Both sides are generated from code with a fixed random seed where needed. Left: unchanged original.py. Right: augmented.py, preserving the same data and chart topology while improving figure aesthetics.

original code renderoriginal codedataset/source imageaugmented code renderaugmented codecomparison imageprocess note

bar · Chart2NCode_0855bef1c82c44309116cd241d54254c

Original code render versus data-preserving FigMirror code augmentation

Both sides are generated from code with a fixed random seed where needed. Left: unchanged original.py. Right: augmented.py, preserving the same data and chart topology while improving figure aesthetics.

original code renderoriginal codedataset/source imageaugmented code renderaugmented codecomparison imageprocess note