ALOcc: Adaptive Lifting-based 3D Semantic Occupancy and Cost Volume-based Flow Prediction

Vision-based semantic occupancy and flow prediction plays a crucial role in providing spatiotemporal cues for real-world tasks, such as autonomous driving. Existing methods prioritize higher accuracy to cater to the demands of these tasks. In this work, we strive to improve performance by introducing a series of targeted improvements for 3D semantic occupancy prediction and flow estimation. First, we introduce an occlusion-aware adaptive lifting mechanism with a depth denoising technique to improve the robustness of 2D-to-3D feature transformation and reduce the reliance on depth priors. Second, we strengthen the semantic consistency between 3D features and their original 2D modalities by utilizing shared semantic prototypes to jointly constrain both 2D and 3D features. This is complemented by confidence- and category-based sampling strategies to tackle long-tail challenges in 3D space. To alleviate the feature encoding burden in the joint prediction of semantics and flow, we propose a BEV cost volume-based prediction method that links flow and semantic features through a cost volume and employs a classification-regression supervision scheme to address the varying flow scales in dynamic scenes. Our purely convolutional architecture framework, named ALOcc, achieves an optimal tradeoff between speed and accuracy achieving state-of-the-art results on multiple benchmarks. On Occ3D and training without the camera visible mask, our ALOcc achieves an absolute gain of 2.5\% in terms of RayIoU while operating at a comparable speed compared to the state-of-the-art, using the same input size (256$\times$704) and ResNet-50 backbone. Our method also achieves 2nd place in the CVPR24 Occupancy and Flow Prediction Competition.

Decoupling Fine Detail and Global Geometry for Compressed Depth Map Super-Resolution

Recovering high-quality depth maps from compressed sources has gained significant attention due to the limitations of consumer-grade depth cameras and the bandwidth restrictions during data transmission. However, current methods still suffer from two challenges. First, bit-depth compression produces a uniform depth representation in regions with subtle variations, hindering the recovery of detailed information. Second, densely distributed random noise reduces the accuracy of estimating the global geometric structure of the scene. To address these challenges, we propose a novel framework, termed geometry-decoupled network (GDNet), for compressed depth map super-resolution that decouples the high-quality depth map reconstruction process by handling global and detailed geometric features separately. To be specific, we propose the fine geometry detail encoder (FGDE), which is designed to aggregate fine geometry details in high-resolution low-level image features while simultaneously enriching them with complementary information from low-resolution context-level image features. In addition, we develop the global geometry encoder (GGE) that aims at suppressing noise and extracting global geometric information effectively via constructing compact feature representation in a low-rank space. We conduct experiments on multiple benchmark datasets, demonstrating that our GDNet significantly outperforms current methods in terms of geometric consistency and detail recovery. In the ECCV 2024 AIM Compressed Depth Upsampling Challenge, our solution won the 1st place award. Our codes will be available.

Rethinking Visual Dependency in Long-Context Reasoning for Large Vision-Language Models

Large Vision-Language Models (LVLMs) excel in cross-model tasks but experience performance declines in long-context reasoning due to overreliance on textual information and reduced visual dependency. In this study, we empirically analyze LVLMs in long-context reasoning, revealing that increased context length leads to a higher dependence on language at the expense of visual dependency. To address this issue, we propose a novel training-free context pruning method that selectively removes less critical textual information. Our approach enhances visual dependency and reduces textual noise, thereby improving LVLM performance in long-context reasoning. We validate our method by constructing a long-context dataset, demonstrating its effectiveness across various LVLMs. Moreover, further analysis confirms the robustness of different token pruning strategies and preliminary explores scaling laws between pruning rates and context length.

High-Precision Self-Supervised Monocular Depth Estimation with Rich-Resource Prior

In the area of self-supervised monocular depth estimation, models that utilize rich-resource inputs, such as high-resolution and multi-frame inputs, typically achieve better performance than models that use ordinary single image input. However, these rich-resource inputs may not always be available, limiting the applicability of these methods in general scenarios. In this paper, we propose Rich-resource Prior Depth estimator (RPrDepth), which only requires single input image during the inference phase but can still produce highly accurate depth estimations comparable to rich resource based methods. Specifically, we treat rich-resource data as prior information and extract features from it as reference features in an offline manner. When estimating the depth for a single-image image, we search for similar pixels from the rich-resource features and use them as prior information to estimate the depth. Experimental results demonstrate that our model outperform other single-image model and can achieve comparable or even better performance than models with rich-resource inputs, only using low-resolution single-image input.

RepVF: A Unified Vector Fields Representation for Multi-task 3D Perception

Concurrent processing of multiple autonomous driving 3D perception tasks within the same spatiotemporal scene poses a significant challenge, in particular due to the computational inefficiencies and feature competition between tasks when using traditional multi-task learning approaches. This paper addresses these issues by proposing a novel unified representation, RepVF, which harmonizes the representation of various perception tasks such as 3D object detection and 3D lane detection within a single framework. RepVF characterizes the structure of different targets in the scene through a vector field, enabling a single-head, multi-task learning model that significantly reduces computational redundancy and feature competition. Building upon RepVF, we introduce RFTR, a network designed to exploit the inherent connections between different tasks by utilizing a hierarchical structure of queries that implicitly model the relationships both between and within tasks. This approach eliminates the need for task-specific heads and parameters, fundamentally reducing the conflicts inherent in traditional multi-task learning paradigms. We validate our approach by combining labels from the OpenLane dataset with the Waymo Open dataset. Our work presents a significant advancement in the efficiency and effectiveness of multi-task perception in autonomous driving, offering a new perspective on handling multiple 3D perception tasks synchronously and in parallel. The code will be available at: https://github.com/jbji/RepVF

AdaOcc: Adaptive Forward View Transformation and Flow Modeling for 3D Occupancy and Flow Prediction

In this technical report, we present our solution for the Vision-Centric 3D Occupancy and Flow Prediction track in the nuScenes Open-Occ Dataset Challenge at CVPR 2024. Our innovative approach involves a dual-stage framework that enhances 3D occupancy and flow predictions by incorporating adaptive forward view transformation and flow modeling. Initially, we independently train the occupancy model, followed by flow prediction using sequential frame integration. Our method combines regression with classification to address scale variations in different scenes, and leverages predicted flow to warp current voxel features to future frames, guided by future frame ground truth. Experimental results on the nuScenes dataset demonstrate significant improvements in accuracy and robustness, showcasing the effectiveness of our approach in real-world scenarios. Our single model based on Swin-Base ranks second on the public leaderboard, validating the potential of our method in advancing autonomous car perception systems.

Multi-threshold Deep Metric Learning for Facial Expression Recognition

Effective expression feature representations generated by a triplet-based deep metric learning are highly advantageous for facial expression recognition (FER). The performance of triplet-based deep metric learning is contingent upon identifying the best threshold for triplet loss. Threshold validation, however, is tough and challenging, as the ideal threshold changes among datasets and even across classes within the same dataset. In this paper, we present the multi-threshold deep metric learning technique, which not only avoids the difficult threshold validation but also vastly increases the capacity of triplet loss learning to construct expression feature representations. We find that each threshold of the triplet loss intrinsically determines a distinctive distribution of inter-class variations and corresponds, thus, to a unique expression feature representation. Therefore, rather than selecting a single optimal threshold from a valid threshold range, we thoroughly sample thresholds across the range, allowing the representation characteristics manifested by thresholds within the range to be fully extracted and leveraged for FER. To realize this approach, we partition the embedding layer of the deep metric learning network into a collection of slices and model training these embedding slices as an end-to-end multi-threshold deep metric learning problem. Each embedding slice corresponds to a sample threshold and is learned by enforcing the corresponding triplet loss, yielding a set of distinct expression features, one for each embedding slice. It makes the embedding layer, which is composed of a set of slices, a more informative and discriminative feature, hence enhancing the FER accuracy. Extensive evaluations demonstrate the superior performance of the proposed approach on both posed and spontaneous facial expression datasets.

Is a 3D-Tokenized LLM the Key to Reliable Autonomous Driving?

Rapid advancements in Autonomous Driving (AD) tasks turned a significant shift toward end-to-end fashion, particularly in the utilization of vision-language models (VLMs) that integrate robust logical reasoning and cognitive abilities to enable comprehensive end-to-end planning. However, these VLM-based approaches tend to integrate 2D vision tokenizers and a large language model (LLM) for ego-car planning, which lack 3D geometric priors as a cornerstone of reliable planning. Naturally, this observation raises a critical concern: Can a 2D-tokenized LLM accurately perceive the 3D environment? Our evaluation of current VLM-based methods across 3D object detection, vectorized map construction, and environmental caption suggests that the answer is, unfortunately, NO. In other words, 2D-tokenized LLM fails to provide reliable autonomous driving. In response, we introduce DETR-style 3D perceptrons as 3D tokenizers, which connect LLM with a one-layer linear projector. This simple yet elegant strategy, termed Atlas, harnesses the inherent priors of the 3D physical world, enabling it to simultaneously process high-resolution multi-view images and employ spatiotemporal modeling. Despite its simplicity, Atlas demonstrates superior performance in both 3D detection and ego planning tasks on nuScenes dataset, proving that 3D-tokenized LLM is the key to reliable autonomous driving. The code and datasets will be released.

IS-Fusion: Instance-Scene Collaborative Fusion for Multimodal 3D Object Detection

Bird's eye view (BEV) representation has emerged as a dominant solution for describing 3D space in autonomous driving scenarios. However, objects in the BEV representation typically exhibit small sizes, and the associated point cloud context is inherently sparse, which leads to great challenges for reliable 3D perception. In this paper, we propose IS-Fusion, an innovative multimodal fusion framework that jointly captures the Instance- and Scene-level contextual information. IS-Fusion essentially differs from existing approaches that only focus on the BEV scene-level fusion by explicitly incorporating instance-level multimodal information, thus facilitating the instance-centric tasks like 3D object detection. It comprises a Hierarchical Scene Fusion (HSF) module and an Instance-Guided Fusion (IGF) module. HSF applies Point-to-Grid and Grid-to-Region transformers to capture the multimodal scene context at different granularities. IGF mines instance candidates, explores their relationships, and aggregates the local multimodal context for each instance. These instances then serve as guidance to enhance the scene feature and yield an instance-aware BEV representation. On the challenging nuScenes benchmark, IS-Fusion outperforms all the published multimodal works to date. Code is available at: https://github.com/yinjunbo/IS-Fusion.

Visual In-Context Learning for Large Vision-Language Models

In Large Visual Language Models (LVLMs), the efficacy of In-Context Learning (ICL) remains limited by challenges in cross-modal interactions and representation disparities. To overcome these challenges, we introduce a novel Visual In-Context Learning (VICL) method comprising Visual Demonstration Retrieval, Intent-Oriented Image Summarization, and Intent-Oriented Demonstration Composition. Our approach retrieves images via ''Retrieval & Rerank'' paradigm, summarises images with task intent and task-specific visual parsing, and composes language-based demonstrations that reduce token count and alleviate cross-modal interaction problem. Experimental evaluations on five visual reasoning datasets demonstrate the effectiveness of our method. Moreover, our extensive experiments leverage information flow analysis to elucidate the effectiveness of our method, and investigate the impact of length and position of demonstrations for LVLM. The use of in-context unlearning further shows promise in resetting specific model knowledge without retraining.