REVIEW 5 cited by
Time Will Tell: New Outlooks and A Baseline for Temporal Multi-View 3D Object Detection
Not yet reviewed by Pith; the record is open.
This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.
SPECIMEN: schema-true, not a live event
T0 review · schema-true
One-sentence machine reading of the paper's core claim.
pith:XXXXXXXX · record.json · timestamp
abstract
While recent camera-only 3D detection methods leverage multiple timesteps, the limited history they use significantly hampers the extent to which temporal fusion can improve object perception. Observing that existing works' fusion of multi-frame images are instances of temporal stereo matching, we find that performance is hindered by the interplay between 1) the low granularity of matching resolution and 2) the sub-optimal multi-view setup produced by limited history usage. Our theoretical and empirical analysis demonstrates that the optimal temporal difference between views varies significantly for different pixels and depths, making it necessary to fuse many timesteps over long-term history. Building on our investigation, we propose to generate a cost volume from a long history of image observations, compensating for the coarse but efficient matching resolution with a more optimal multi-view matching setup. Further, we augment the per-frame monocular depth predictions used for long-term, coarse matching with short-term, fine-grained matching and find that long and short term temporal fusion are highly complementary. While maintaining high efficiency, our framework sets new state-of-the-art on nuScenes, achieving first place on the test set and outperforming previous best art by 5.2% mAP and 3.7% NDS on the validation set. Code will be released $\href{https://github.com/Divadi/SOLOFusion}{here.}$
Forward citations
Cited by 5 Pith papers
-
FDR-Occ: Factorized Dense Routing for Full-Spectrum 3D Occupancy Prediction
Factorized Dense Routing approximates unconstrained 2D-to-3D feature mixing by hierarchical tensor contractions, yielding global-context occupancy prediction that remains robust without camera extrinsics.
-
MVAT: Multi-View Aware Teacher for Weakly Supervised 3D Object Detection
MVAT aggregates temporally repeated 2D-box-constrained LiDAR views of static objects to produce 3D pseudo-labels, then distills them into a single-frame 3D detector, achieving 47.6 mAP on nuScenes with no 3D labels.
-
GS-Occ3D: Scaling Vision-only Occupancy Reconstruction with Gaussian Splatting
A camera-only Gaussian-surfel pipeline reconstructs full Waymo scenes, converts them to binary occupancy labels, and trains CVT-Occ to generalize on Occ3D-Waymo and Occ3D-nuScenes at a level close to or above LiDAR-la...
-
Progressive Bird's Eye View Perception for Safety-Critical Autonomous Driving: A Comprehensive Survey
A safety-critical survey that organizes BEV perception into single-modality, multimodal, and collaborative stages and consolidates robustness evidence that multimodal fusion degrades far less than single-modality perc...
-
Occupancy Learning with Spatiotemporal Memory
ST-Occ improves 3D occupancy prediction for self-driving by storing a compact scene-level memory of past frames and conditioning current predictions on it with uncertainty-aware attention, gaining 3 mIoU over prior st...
Discussion (0). Continue with ORCID to comment.