AI Paper Atlas

Research map / Label-efficient and robust learning / Self-supervised and contrastive

Differentiable Raycasting for Self-Supervised Occupancy Forecasting

Tarasha Khurana, Peiyun Hu, Achal Dave, Jason Ziglar, David Held, Deva Ramanan

ECCV 2022 · Label-efficient and robust learning › Self-supervised and contrastive · approach: Pretext-task visual pre-training · 68 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 differentiable raycasting to decouple scene and ego motionfreespace forecaster · nuScenes valF1: 0.665 -> 0.777Improved
add differentiable raycasting to decouple scene and ego motionfreespace forecaster · ONCE valF1: 0.635 -> 0.787Improved
add differentiable raycasting (ablation)motion planner · nuScenes valbox collision 3s (%): 1.10 -> 0.88Improved
add differentiable raycasting (ablation)motion planner · nuScenes valL2 error 3s (m): 2.64 -> 2.78Hurt
add multi-task occupancy forecasting auxiliary task (ablation)IL max-margin planner · nuScenes valbox collision 3s (%): 1.95 -> 1.29Improved
add multi-task occupancy forecasting auxiliary task (ablation)IL max-margin planner · nuScenes valL2 error 3s (m): 2.47 -> 2.30No clear effect
add freespace-guided cost margin (ablation)IL max-margin planner · nuScenes valbox collision 3s (%): 1.95 -> 1.07Improved
add freespace-guided cost margin (ablation)IL max-margin planner · nuScenes valL2 error 3s (m): 2.47 -> 2.54Hurt
add freespace-guided cost margin to multi-task planner (ablation)multi-task planner · nuScenes valbox collision 3s (%): 1.29 -> 1.10Improved
add differentiable raycasting (ablation)motion planner · ONCE valbox collision 3s (%): 2.61 -> 2.79Hurt
add differentiable raycasting (ablation)motion planner · ONCE valpoint collision 3s (%): 1.10 -> 0.74Improved
increase self-supervised training data from 8K to 94KOurs · ONCE valL2 error 3s (m): 4.43 -> 2.90Improved

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