Search papers, labs, and topics across Lattice.
3
0
3
1
Certified Training outperforms Adversarial Training, achieving over 80% robust accuracy against motion blur while ensuring formal safety guarantees.
Tighter robustness guarantees and 1.7x higher certified accuracy for 3D CNNs in safety-critical applications could redefine verification standards in AI.
Finally, a formal verification method exists that can handle the complex coordinate transformations and IoU metrics inherent in realistic anchor-based object detectors like YOLO and SSD.
