Anonymous submission — under double-blind review. Author and affiliation details are withheld.
← RareOcc project page  ·  λ=0.15 (paper setting)

Occupancy generation vs. UniScene-v2 — λ=0.05

After submission, we found that the released UniScene-v2 code sets the noise-prior weight to λ=0.05 whereas our paper reports λ=0.15. This page re-evaluates UniScene-v2 at λ=0.05. The change is marginal and does not alter any of the paper's contributions or conclusions: ROCC still outperforms UniScene-v2 by a wide margin on every metric and set. We report it for full transparency. Results at the paper's λ=0.15 are here.

UniScene-v2UniScene-v2Dense+ bgROCC
SetMetric(λ=0.05)(λ=0)CNNhead‡(ours)
mini-valfg-mIoU ↑20.1933.5545.6249.25
mIoUUS29.5338.0949.0151.65
occ-IoU ↑22.2424.7424.1927.25
key subsetfg-mIoU ↑20.060.0332.9144.9048.45
mIoUUS29.3713.1237.5148.3550.92
occ-IoU ↑21.657.3123.9023.4026.17

Occupancy generation on two views of the nuPlan-Occ validation split: mini-val (the full validation set) and key subset (one key frame per clip). One shared per-class IoU accumulator scores every column on identical ground truth. UniScene-v2's λ supplies a fraction of the ground-truth occupancy latent of the scored frame; at λ=0 it collapses. ‡ trained at a larger effective batch than the other columns.