AdvIRL uses PPO to adjust Instant-NGP parameters so that CLIP misclassifies rendered 3D objects, with results on banana, truck, horse, and lighthouse scenes.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2024 1verdicts
REJECT 1representative citing papers
citing papers explorer
-
AdvIRL: Reinforcement Learning-Based Adversarial Attacks on 3D NeRF Models
AdvIRL uses PPO to adjust Instant-NGP parameters so that CLIP misclassifies rendered 3D objects, with results on banana, truck, horse, and lighthouse scenes.