CleverDistiller: Simple and Spatially Consistent Cross-modal Distillation

• URL: http://arxiv.org/abs/2503.09878v2
• Date: Mon, 25 Aug 2025 09:34:28 GMT
• Title: CleverDistiller: Simple and Spatially Consistent Cross-modal Distillation
• Authors: Hariprasath Govindarajan, Maciej K. Wozniak, Marvin Klingner, Camille Maurice, B Ravi Kiran, Senthil Yogamani,
• Abstract summary: CleverDistiller is a self-supervised, cross-modal 2D-to-3D KD framework.<n>It achieves state-of-the-art performance in both semantic segmentation and 3D object detection by up to 10% mIoU.
• Score: 7.246959698735599
• License: http://creativecommons.org/licenses/by-nc-nd/4.0/
• Abstract: Vision foundation models (VFMs) such as DINO have led to a paradigm shift in 2D camera-based perception towards extracting generalized features to support many downstream tasks. Recent works introduce self-supervised cross-modal knowledge distillation (KD) as a way to transfer these powerful generalization capabilities into 3D LiDAR-based models. However, they either rely on highly complex distillation losses, pseudo-semantic maps, or limit KD to features useful for semantic segmentation only. In this work, we propose CleverDistiller, a self-supervised, cross-modal 2D-to-3D KD framework introducing a set of simple yet effective design choices: Unlike contrastive approaches relying on complex loss design choices, our method employs a direct feature similarity loss in combination with a multi layer perceptron (MLP) projection head to allow the 3D network to learn complex semantic dependencies throughout the projection. Crucially, our approach does not depend on pseudo-semantic maps, allowing for direct knowledge transfer from a VFM without explicit semantic supervision. Additionally, we introduce the auxiliary self-supervised spatial task of occupancy prediction to enhance the semantic knowledge, obtained from a VFM through KD, with 3D spatial reasoning capabilities. Experiments on standard autonomous driving benchmarks for 2D-to-3D KD demonstrate that CleverDistiller achieves state-of-the-art performance in both semantic segmentation and 3D object detection (3DOD) by up to 10% mIoU, especially when fine tuning on really low data amounts, showing the effectiveness of our simple yet powerful KD strategy

Related papers

• D3D-VLP: Dynamic 3D Vision-Language-Planning Model for Embodied Grounding and Navigation [66.7166217399105]
  Embodied agents face a critical dilemma that end-to-end models lack interpretability and explicit 3D reasoning.<n>Our model introduces two key innovations: 1) A Dynamic 3D Chain-of-Thought (3D CoT) that unifies planning, grounding, navigation, and question answering within a single 3D-VLM and CoT pipeline; 2) A Synergistic Learning from Fragmented Supervision (SLFS) strategy, which uses a masked autoregressive loss to learn from massive and partially-annotated hybrid data.
  arXiv  Detail & Related papers  (2025-12-14T09:53:15Z)
• Unlocking 3D Affordance Segmentation with 2D Semantic Knowledge [45.19482892758984]
  Affordance segmentation aims to parse 3D objects into functionally distinct parts, bridging recognition and interaction for applications in robotic manipulation, embodied AI, and AR.<n>We introduce Cross-Modal Affinity Transfer (CMAT), a pre-training strategy that aligns a 3D encoder with lifted 2D semantics and jointly optimize reconstruction, affinity, and diversity to yield semantically organized representations.<n>We further design the Cross-modal Affordance Transformer (CAST), which integrates multi-modal prompts with CMAT-pretrained features to generate precise, prompt-aware segmentation maps.
  arXiv  Detail & Related papers  (2025-10-09T15:01:26Z)
• Domain Adaptation-Based Crossmodal Knowledge Distillation for 3D Semantic Segmentation [3.6903284283843942]
  Traditional approaches rely on extensive data for annotated point cloud analysis.<n>Realworld image datasets offer abundant availability and substantial scale.<n>We propose two crossmodal knowledge distillation methods: Unsupervised Domain Knowledge Distillation (UDAKD) and Feature and Semantic-based Knowledge Distillation (FSKD)
  arXiv  Detail & Related papers  (2025-08-30T06:34:39Z)
• CMF-IoU: Multi-Stage Cross-Modal Fusion 3D Object Detection with IoU Joint Prediction [29.7092783661859]
  Multi-modal methods based on camera and LiDAR sensors have garnered significant attention in the field of 3D detection.<n>We introduce a multi-stage cross-modal fusion 3D detection framework, termed CMF-IOU, to address the challenge of aligning 3D spatial and 2D semantic information.
  arXiv  Detail & Related papers  (2025-08-18T13:32:07Z)
• Learning A Zero-shot Occupancy Network from Vision Foundation Models via Self-supervised Adaptation [41.98740330990215]
  This work proposes a novel approach that bridges 2D vision foundation models with 3D tasks. We leverage the zero-shot capabilities of vision-language models for image semantics. We project the semantics into 3D space using the reconstructed metric depth, thereby providing 3D supervision.
  arXiv  Detail & Related papers  (2025-03-10T09:54:40Z)
• A Lesson in Splats: Teacher-Guided Diffusion for 3D Gaussian Splats Generation with 2D Supervision [65.33043028101471]
  We introduce a diffusion model for Gaussian Splats, SplatDiffusion, to enable generation of three-dimensional structures from single images.<n>Existing methods rely on deterministic, feed-forward predictions, which limit their ability to handle the inherent ambiguity of 3D inference from 2D data.
  arXiv  Detail & Related papers  (2024-12-01T00:29:57Z)
• UltimateDO: An Efficient Framework to Marry Occupancy Prediction with 3D Object Detection via Channel2height [2.975860548186652]
  Occupancy and 3D object detection are two standard tasks in modern autonomous driving system. We propose a method to achieve fast 3D object detection and occupancy prediction (UltimateDO)
  arXiv  Detail & Related papers  (2024-09-17T13:14:13Z)
• Semi-supervised 3D Semantic Scene Completion with 2D Vision Foundation Model Guidance [8.07701188057789]
  We introduce a novel semi-supervised framework to alleviate the dependency on densely annotated data.<n>Our approach leverages 2D foundation models to generate essential 3D scene geometric and semantic cues.<n>Our method achieves up to 85% of the fully-supervised performance using only 10% labeled data.
  arXiv  Detail & Related papers  (2024-08-21T12:13:18Z)
• Weakly Supervised Monocular 3D Detection with a Single-View Image [58.57978772009438]
  Monocular 3D detection aims for precise 3D object localization from a single-view image. We propose SKD-WM3D, a weakly supervised monocular 3D detection framework. We show that SKD-WM3D surpasses the state-of-the-art clearly and is even on par with many fully supervised methods.
  arXiv  Detail & Related papers  (2024-02-29T13:26:47Z)
• ODM3D: Alleviating Foreground Sparsity for Semi-Supervised Monocular 3D Object Detection [15.204935788297226]
  ODM3D framework entails cross-modal knowledge distillation at various levels to inject LiDAR-domain knowledge into a monocular detector during training. By identifying foreground sparsity as the main culprit behind existing methods' suboptimal training, we exploit the precise localisation information embedded in LiDAR points. Our method ranks 1st in both KITTI validation and test benchmarks, significantly surpassing all existing monocular methods, supervised or semi-supervised.
  arXiv  Detail & Related papers  (2023-10-28T07:12:09Z)
• Weakly Supervised Monocular 3D Object Detection using Multi-View Projection and Direction Consistency [78.76508318592552]
  Monocular 3D object detection has become a mainstream approach in automatic driving for its easy application. Most current methods still rely on 3D point cloud data for labeling the ground truths used in the training phase. We propose a new weakly supervised monocular 3D objection detection method, which can train the model with only 2D labels marked on images.
  arXiv  Detail & Related papers  (2023-03-15T15:14:00Z)
• Joint-MAE: 2D-3D Joint Masked Autoencoders for 3D Point Cloud Pre-training [65.75399500494343]
  Masked Autoencoders (MAE) have shown promising performance in self-supervised learning for 2D and 3D computer vision. We propose Joint-MAE, a 2D-3D joint MAE framework for self-supervised 3D point cloud pre-training.
  arXiv  Detail & Related papers  (2023-02-27T17:56:18Z)
• Homography Loss for Monocular 3D Object Detection [54.04870007473932]
  A differentiable loss function, termed as Homography Loss, is proposed to achieve the goal, which exploits both 2D and 3D information. Our method yields the best performance compared with the other state-of-the-arts by a large margin on KITTI 3D datasets.
  arXiv  Detail & Related papers  (2022-04-02T03:48:03Z)
• The Devil is in the Task: Exploiting Reciprocal Appearance-Localization Features for Monocular 3D Object Detection [62.1185839286255]
  Low-cost monocular 3D object detection plays a fundamental role in autonomous driving. We introduce a Dynamic Feature Reflecting Network, named DFR-Net. We rank 1st among all the monocular 3D object detectors in the KITTI test set.
  arXiv  Detail & Related papers  (2021-12-28T07:31:18Z)
• Cylindrical and Asymmetrical 3D Convolution Networks for LiDAR-based Perception [122.53774221136193]
  State-of-the-art methods for driving-scene LiDAR-based perception often project the point clouds to 2D space and then process them via 2D convolution. A natural remedy is to utilize the 3D voxelization and 3D convolution network. We propose a new framework for the outdoor LiDAR segmentation, where cylindrical partition and asymmetrical 3D convolution networks are designed to explore the 3D geometric pattern.
  arXiv  Detail & Related papers  (2021-09-12T06:25:11Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.