Browse Teach Curate

Active Learning Test Input Validation for Vision DL Systems

HiL-TV: automated validation of synthetic test inputs for vision-based deep learning systems. A human-in-the-loop active learning loop trains an ML classifier over multiple image-comparison metrics (PSNR, SSIM, VIF, VAE reconstruction error, semantic segmentation score) to separate valid from invalid generated test images, balancing accuracy against manual labeling effort and yielding Pareto-optimal accuracy-effort trade-offs on an industrial and a public dataset (ICSE 2025 SEIP; arXiv:2501.01606).

Johnny Five Text

Test Input Validation for Vision-based DL Systems: An Active Learning Approach

Testing deep learning (DL) systems requires extensive and diverse, yet valid, test inputs. While synthetic test input generation methods, such as metamorphic testing, are widely used for DL testing, …