Synthetic images from Unreal Engine can replace up to 80-90% of real training images in two binary image classification tasks, with a small accuracy trade-off and improved train-test generalization.
Searching for mobilenetv3
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.CV 1years
2025 1verdicts
CONDITIONAL 1roles
background 1polarities
unclear 1representative citing papers
citing papers explorer
-
The effects of using created synthetic images in computer vision training
Synthetic images from Unreal Engine can replace up to 80-90% of real training images in two binary image classification tasks, with a small accuracy trade-off and improved train-test generalization.