AI Paper Atlas

Research map / Federated learning and privacy / Privacy attacks and secure learning

Private Multiparty Perception for Navigation

Hui Lu, Mia Chiquier, Carl Vondrick

NeurIPS 2022 · Federated learning and privacy › Privacy attacks and secure learning · approach: Visual privacy and face de-identification · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Add multiview camera inputs and perform inference with two-party MPCCipherNav navigation model · Obstacle World test, Overallaccuracy: 74.0% -> 96.6%Improved
Add multiview camera inputs and perform inference with five-party MPCCipherNav navigation model · Obstacle World test, Overallaccuracy: 74.0% -> 96.9%Improved
Replace plaintext multiview inference with two-party MPCNavigation with camera views · Obstacle World test, Overallaccuracy: 97.1% -> 96.6%Hurt
Replace plaintext multiview inference with five-party MPCNavigation with camera views · Obstacle World test, Overallaccuracy: 97.1% -> 96.9%Hurt
Use a deterministic rather than random starting view directionFirst Person baseline · Obstacle World test, Overallaccuracy: 72.0% -> 88.8%Improved
Add external camera views to map and agent inputsNavigation model · Obstacle World successful paths, Overallpath efficiency: 95.5 % -> 99.5 %Improved
Use two-party MPC instead of plaintext for multiview inferenceNavigation with camera views · Obstacle World successful paths, Overallpath efficiency: 99.4 % -> 99.5 %Mixed
Use five-party MPC instead of plaintext for multiview inferenceNavigation with camera views · Obstacle World successful paths, Overallpath efficiency: 99.4 % -> 99.2 %Hurt
Perform inference with two-party MPC rather than plaintextCipherNav action prediction network · Obstacle World runtime, batch size 100, GPUruntime (seconds): 0.00028 -> 0.15Hurt
Perform inference with five-party MPC rather than plaintextCipherNav action prediction network · Obstacle World runtime, batch size 100, GPUruntime (seconds): 0.00028 -> 0.42Hurt
Use encrypted rather than plaintext camera featuresScene reconstruction classifier · Obstacle World scene reconstructionaccuracy: 93.6% -> 49.7%Hurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy