mlsanityData to 7 Oct 2026

FQ-PETR: Fully Quantized Position Embedding Transformation for Multi-View 3D Object Detection

arXiv 2511.09347 ↗First posted 12 Nov 2025cs.CVcited by 0

Not on the map. Why?

Abstract (arXiv)

Camera-based multi-view 3D detection is crucial for autonomous driving. PETR and its variants (PETRs) excel in benchmarks but face deployment challenges due to high computational cost and memory footprint. Quantization is an effective technique for compressing deep neural networks by reducing the bit width of weights and activations. However, directly applying existing quantization methods to PETRs leads to severe accuracy degradation. This issue primarily arises from two key challenges: (1) significant magnitude disparity between multi-modal features-specifically, image features and camera-ray positional embeddings (PE), and (2) the inefficiency and approximation error of quantizing non-linear operators, which commonly rely on hardware-unfriendly computations. In this paper, we propose FQ-PETR, a fully quantized framework for PETRs, featuring three key innovations: (1) Quantization-Friendly LiDAR-ray Position Embedding (QFPE): Replacing multi-point sampling with LiDAR-prior-guided single-point sampling and anchor-based embedding eliminates problematic non-linearities (e.g., inverse-sigmoid) and aligns PE scale with image features, preserving accuracy. (2) Dual-Lookup Table (DULUT): This algorithm approximates complex non-linear functions using two cascaded linear LUTs, achieving high fidelity with minimal entries and no specialized hardware. (3) Quantization After Numerical Stabilization (QANS): Performing quantization after softmax numerical stabilization mitigates attention distortion from large inputs. On PETRs (e.g. PETR, StreamPETR, PETRv2, MV2d), FQ-PETR under W8A8 achieves near-floating-point accuracy (1% degradation) while reducing latency by up to 75%, significantly outperforming existing PTQ and QAT baselines.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Builds on 36 of its 68 references that are in the map. First posted on arXiv 12 Nov 2025.

Builds on

36 of 68 references.

References · 36

  1. 2023FlashAttention-2: Faster Attention with Better Parallelism and Work Partitioningcited by 2,551
  2. 2022BEVFormer: Learning Bird's-Eye-View Representation from Multi-Camera Images via Spatiotemporal Transformerscited by 1,213
  3. 2020Center-based 3D Object Detection and Trackingcited by 1,143
  4. 2020Lift, Splat, Shoot: Encoding Images From Arbitrary Camera Rigs by Implicitly Unprojecting to 3Dcited by 1,040
  5. 2021BEVDet: High-performance Multi-camera 3D Object Detection in Bird-Eye-Viewcited by 711
  6. 2019Learned Step Size Quantizationcited by 604
  7. 2020Up or Down? Adaptive Rounding for Post-Training Quantizationcited by 600
  8. 2024QuaRot: Outlier-Free 4-Bit Inference in Rotated LLMscited by 598
  9. 2021BRECQ: Pushing the Limit of Post-Training Quantization by Block Reconstructioncited by 464
  10. 2024SpinQuant: LLM quantization with learned rotationscited by 446
  11. 2023OmniQuant: Omnidirectionally Calibrated Quantization for Large Language Modelscited by 403
  12. 2022PETRv2: A Unified Framework for 3D Perception from Multi-Camera Imagescited by 351
24 more
  1. 2019Data-Free Quantization Through Weight Equalization and Bias Correctioncited by 328
  2. 2018LQ-Nets: Learned Quantization for Highly Accurate and Compact Deep Neural Networkscited by 272
  3. 2021I-BERT: Integer-only BERT Quantizationcited by 250
  4. 2019Low-bit Quantization of Neural Networks for Efficient Inferencecited by 185
  5. 2020LSQ+: Improving low-bit quantization through learnable offsets and better initializationcited by 180
  6. 2022Outlier Suppression: Pushing the Limit of Low-bit Transformer Language Modelscited by 176
  7. 2021PTQ4ViT: Post-training quantization for vision transformers with twin uniform quantizationcited by 173
  8. 2021FQ-ViT: Post-Training Quantization for Fully Quantized Vision Transformercited by 148
  9. 2022RepQ-ViT: Scale Reparameterization for Post-Training Quantization of Vision Transformerscited by 134
  10. 2025OstQuant: Refining Large Language Model Quantization with Orthogonal and Scaling Transformations for Better Distribution Fittingcited by 94
  11. 2022I-ViT: Integer-only Quantization for Efficient Vision Transformer Inferencecited by 93
  12. 2023Object as Query: Lifting any 2D Object Detector to 3D Detectioncited by 38
  13. 2025MQuant: Unleashing the Inference Potential of Multimodal Large Language Models via Full Static Quantizationcited by 35
  14. 2025MoEQuant: Enhancing Quantization for Mixture-of-Experts Large Language Models via Expert-Balanced Sampling and Affinity Guidancecited by 32
  15. 2025SepPrune: Structured Pruning for Efficient Deep Speech Separationcited by 31
  16. 2021NN-LUT: Neural Approximation of Non-Linear Operations for Efficient Transformer Inferencecited by 25
  17. 2025FedSWA: Improving Generalization in Federated Learning with Highly Heterogeneous Data via Momentum-Based Stochastic Controlled Weight Averagingcited by 25
  18. 2023Q-DETR: An Efficient Low-Bit Quantized Detection Transformercited by 23
  19. 2025Knowledge-Aligned Counterfactual-Enhancement Diffusion Perception for Unsupervised Cross-Domain Visual Emotion Recognitioncited by 16
  20. 20223DPPE: 3D Point Positional Encoding for Multi-Camera 3D Object Detection Transformerscited by 15
  21. 2024Information Entropy Guided Height-aware Histogram for Quantization-friendly Pillar Feature Encodercited by 14
  22. 2024RayFormer: Improving Query-Based Multi-Camera 3D Object Detection via Ray-Centric Strategiescited by 9
  23. 2023Post-Training Quantization for Re-parameterization via Coarse & Fine Weight Splittingcited by 6
  24. 2024P2-ViT: Power-of-Two Post-Training Quantization and Acceleration for Fully Quantized Vision Transformercited by 4

Built on by

Not cited yet.