Decision-FLIP reduces model size and speeds Gaussian occupancy predictors
Decision-FLIP: Budget-Free Structured Pruning for Gaussian Occupancy Models
Computer Vision and Pattern Recognition
Summary
Gaussian occupancy models use many components to decide what parts of space are occupied, but these components are often fixed in size and can be inefficient. The authors introduce Decision-FLIP, a new way to automatically identify which parts of the model can be removed without needing labeled data. Their technique measures how removing pieces changes the model’s decisions, and then compiles a smaller, faster model accordingly. This method saves computation and memory while keeping accuracy nearly the same.
What this means in practice
- •For computer vision engineers: Reduce latency and memory use of occupancy predictor models without labeled data by automatically pruning unnecessary Gaussian components.
- •For autonomous vehicle teams: Improve real-time 3D scene understanding by streamlining Gaussian occupancy banks, leading to faster and more efficient spatial sensing.
Authors
Jingxing Li, Anurag Paul, Subhasis Das, Cindy Xu
Abstract
Occupancy predictors rely on large persistent banks of Gaussians or learned queries whose size is fixed by hand. Common pruning proxies, such as opacity, activation frequency, or parameter magnitude, do not measure whether removing a row changes the model's voxel decisions. We introduce Decision-FLIP, a budget-free compilation framework whose scoring and structure selection need only unlabeled calibration inputs. Exact Mean-FLIP scores each persistent row by the number of voxel argmax decisions that change when the row is removed from the readout; a guarded Otsu split of the log-score distribution selects the survivor count without a target and abstains when the split is unsupported; and the survivors are physically compiled into a smaller checkpoint, so camera sampling, attention, and aggregation shrink with the bank. For streaming banks, scores are integrated over each row's lifecycle through warp, death, and refill. Across four Gaussian-bank model-dataset settings and one persistent-query transfer, the automatically selected cuts remove 12-79% of eligible rows and cut bank/query-dependent latency by 7-45% across settings with component profiles, with whole-model savings of up to 8.0% on the primary endpoints bounded by fixed-cost image, voxel, and sensor modules and peak-memory savings of up to 16.6%. Under matched one-epoch recovery controls, semantic mIoU changes by at most 0.18 point. Same-$K$ controls show that the ranking, not only the bank size, matters: random removal collapses GaussianFormer-1 and OccFormer, while label-free Mean-FLIP matches or outperforms supervised Taylor ranking on all five settings where that control was run. The method applies when removable primitives keep a stable checkpoint-row identity; it is budget-free in survivor count, not parameter-free.