Jiajun Deng

Yingjie Wang, Jiajun Deng, Yao Li et al.

LiDAR and Radar are two complementary sensing approaches in that LiDAR specializes in capturing an object's 3D shape while Radar provides longer detection ranges as well as velocity hints. Though seemingly natural, how to efficiently combine them for improved feature representation is still unclear. The main challenge arises from that Radar data are extremely sparse and lack height information. Therefore, directly integrating Radar features into LiDAR-centric detection networks is not optimal. In this work, we introduce a bi-directional LiDAR-Radar fusion framework, termed Bi-LRFusion, to tackle the challenges and improve 3D detection for dynamic objects. Technically, Bi-LRFusion involves two steps: first, it enriches Radar's local features by learning important details from the LiDAR branch to alleviate the problems caused by the absence of height information and extreme sparsity; second, it combines LiDAR features with the enhanced Radar features in a unified bird's-eye-view representation. We conduct extensive experiments on nuScenes and ORR datasets, and show that our Bi-LRFusion achieves state-of-the-art performance for detecting dynamic objects. Notably, Radar data in these two datasets have different formats, which demonstrates the generalizability of our method. Codes are available at https://github.com/JessieW0806/BiLRFusion.

Yunyao Mao, Wengang Zhou, Zhenbo Lu et al.

In 3D action recognition, there exists rich complementary information between skeleton modalities. Nevertheless, how to model and utilize this information remains a challenging problem for self-supervised 3D action representation learning. In this work, we formulate the cross-modal interaction as a bidirectional knowledge distillation problem. Different from classic distillation solutions that transfer the knowledge of a fixed and pre-trained teacher to the student, in this work, the knowledge is continuously updated and bidirectionally distilled between modalities. To this end, we propose a new Cross-modal Mutual Distillation (CMD) framework with the following designs. On the one hand, the neighboring similarity distribution is introduced to model the knowledge learned in each modality, where the relational information is naturally suitable for the contrastive frameworks. On the other hand, asymmetrical configurations are used for teacher and student to stabilize the distillation process and to transfer high-confidence information between modalities. By derivation, we find that the cross-modal positive mining in previous works can be regarded as a degenerated version of our CMD. We perform extensive experiments on NTU RGB+D 60, NTU RGB+D 120, and PKU-MMD II datasets. Our approach outperforms existing self-supervised methods and sets a series of new records. The code is available at: https://github.com/maoyunyao/CMD

Hao Feng, Wengang Zhou, Jiajun Deng et al.

In document image rectification, there exist rich geometric constraints between the distorted image and the ground truth one. However, such geometric constraints are largely ignored in existing advanced solutions, which limits the rectification performance. To this end, we present DocGeoNet for document image rectification by introducing explicit geometric representation. Technically, two typical attributes of the document image are involved in the proposed geometric representation learning, i.e., 3D shape and textlines. Our motivation arises from the insight that 3D shape provides global unwarping cues for rectifying a distorted document image while overlooking the local structure. On the other hand, textlines complementarily provide explicit geometric constraints for local patterns. The learned geometric representation effectively bridges the distorted image and the ground truth one. Extensive experiments show the effectiveness of our framework and demonstrate the superiority of our DocGeoNet over state-of-the-art methods on both the DocUNet Benchmark dataset and our proposed DIR300 test set. The code is available at https://github.com/fh2019ustc/DocGeoNet.

Hao Feng, Shaokai Liu, Jiajun Deng et al.

In recent years, tremendous efforts have been made on document image rectification, but existing advanced algorithms are limited to processing restricted document images, i.e., the input images must incorporate a complete document. Once the captured image merely involves a local text region, its rectification quality is degraded and unsatisfactory. Our previously proposed DocTr, a transformer-assisted network for document image rectification, also suffers from this limitation. In this work, we present DocTr++, a novel unified framework for document image rectification, without any restrictions on the input distorted images. Our major technical improvements can be concluded in three aspects. Firstly, we upgrade the original architecture by adopting a hierarchical encoder-decoder structure for multi-scale representation extraction and parsing. Secondly, we reformulate the pixel-wise mapping relationship between the unrestricted distorted document images and the distortion-free counterparts. The obtained data is used to train our DocTr++ for unrestricted document image rectification. Thirdly, we contribute a real-world test set and metrics applicable for evaluating the rectification quality. To our best knowledge, this is the first learning-based method for the rectification of unrestricted document images. Extensive experiments are conducted, and the results demonstrate the effectiveness and superiority of our method. We hope our DocTr++ will serve as a strong baseline for generic document image rectification, prompting the further advancement and application of learning-based algorithms. The source code and the proposed dataset are publicly available at https://github.com/fh2019ustc/DocTr-Plus.

Hao Feng, Keyi Zhou, Wengang Zhou et al.

Contour-based instance segmentation has been actively studied, thanks to its flexibility and elegance in processing visual objects within complex backgrounds. In this work, we propose a novel deep network architecture, i.e., PolySnake, for generic contour-based instance segmentation. Motivated by the classic Snake algorithm, the proposed PolySnake achieves superior and robust segmentation performance with an iterative and progressive contour refinement strategy. Technically, PolySnake introduces a recurrent update operator to estimate the object contour iteratively. It maintains a single estimate of the contour that is progressively deformed toward the object boundary. At each iteration, PolySnake builds a semantic-rich representation for the current contour and feeds it to the recurrent operator for further contour adjustment. Through the iterative refinements, the contour progressively converges to a stable status that tightly encloses the object instance. Beyond the scope of general instance segmentation, extensive experiments are conducted to validate the effectiveness and generalizability of our PolySnake in two additional specific task scenarios, including scene text detection and lane detection. The results demonstrate that the proposed PolySnake outperforms the existing advanced methods on several multiple prevalent benchmarks across the three tasks. The codes and pre-trained models are available at https://github.com/fh2019ustc/PolySnake

Yufei Yin, Jiajun Deng, Wengang Zhou et al.

Recent progress in weakly supervised object detection is featured by a combination of multiple instance detection networks (MIDN) and ordinal online refinement. However, with only image-level annotation, MIDN inevitably assigns high scores to some unexpected region proposals when generating pseudo labels. These inaccurate high-scoring region proposals will mislead the training of subsequent refinement modules and thus hamper the detection performance. In this work, we explore how to ameliorate the quality of pseudo-labeling in MIDN. Formally, we devise Cyclic-Bootstrap Labeling (CBL), a novel weakly supervised object detection pipeline, which optimizes MIDN with rank information from a reliable teacher network. Specifically, we obtain this teacher network by introducing a weighted exponential moving average strategy to take advantage of various refinement modules. A novel class-specific ranking distillation algorithm is proposed to leverage the output of weighted ensembled teacher network for distilling MIDN with rank information. As a result, MIDN is guided to assign higher scores to accurate proposals among their neighboring ones, thus benefiting the subsequent pseudo labeling. Extensive experiments on the prevalent PASCAL VOC 2007 \& 2012 and COCO datasets demonstrate the superior performance of our CBL framework. Code will be available at https://github.com/Yinyf0804/WSOD-CBL/.

Jiajun Deng, Zhengyuan Yang, Daqing Liu et al.

In this work, we explore neat yet effective Transformer-based frameworks for visual grounding. The previous methods generally address the core problem of visual grounding, i.e., multi-modal fusion and reasoning, with manually-designed mechanisms. Such heuristic designs are not only complicated but also make models easily overfit specific data distributions. To avoid this, we first propose TransVG, which establishes multi-modal correspondences by Transformers and localizes referred regions by directly regressing box coordinates. We empirically show that complicated fusion modules can be replaced by a simple stack of Transformer encoder layers with higher performance. However, the core fusion Transformer in TransVG is stand-alone against uni-modal encoders, and thus should be trained from scratch on limited visual grounding data, which makes it hard to be optimized and leads to sub-optimal performance. To this end, we further introduce TransVG++ to make two-fold improvements. For one thing, we upgrade our framework to a purely Transformer-based one by leveraging Vision Transformer (ViT) for vision feature encoding. For another, we devise Language Conditioned Vision Transformer that removes external fusion modules and reuses the uni-modal ViT for vision-language fusion at the intermediate layers. We conduct extensive experiments on five prevalent datasets, and report a series of state-of-the-art records.

Hao Feng, Wendi Wang, Jiajun Deng et al.

In fisheye images, rich distinct distortion patterns are regularly distributed in the image plane. These distortion patterns are independent of the visual content and provide informative cues for rectification. To make the best of such rectification cues, we introduce SimFIR, a simple framework for fisheye image rectification based on self-supervised representation learning. Technically, we first split a fisheye image into multiple patches and extract their representations with a Vision Transformer (ViT). To learn fine-grained distortion representations, we then associate different image patches with their specific distortion patterns based on the fisheye model, and further subtly design an innovative unified distortion-aware pretext task for their learning. The transfer performance on the downstream rectification task is remarkably boosted, which verifies the effectiveness of the learned representations. Extensive experiments are conducted, and the quantitative and qualitative results demonstrate the superiority of our method over the state-of-the-art algorithms as well as its strong generalization ability on real-world fisheye images.

Xuanyu Yi, Jiajun Deng, Qianru Sun et al.

We tackle the data scarcity challenge in few-shot point cloud recognition of 3D objects by using a joint prediction from a conventional 3D model and a well-trained 2D model. Surprisingly, such an ensemble, though seems trivial, has hardly been shown effective in recent 2D-3D models. We find out the crux is the less effective training for the ''joint hard samples'', which have high confidence prediction on different wrong labels, implying that the 2D and 3D models do not collaborate well. To this end, our proposed invariant training strategy, called InvJoint, does not only emphasize the training more on the hard samples, but also seeks the invariance between the conflicting 2D and 3D ambiguous predictions. InvJoint can learn more collaborative 2D and 3D representations for better ensemble. Extensive experiments on 3D shape classification with widely adopted ModelNet10/40, ScanObjectNN and Toys4K, and shape retrieval with ShapeNet-Core validate the superiority of our InvJoint.

Shujie Hu, Xurong Xie, Zengrui Jin et al.

Automatic recognition of disordered and elderly speech remains a highly challenging task to date due to the difficulty in collecting such data in large quantities. This paper explores a series of approaches to integrate domain adapted SSL pre-trained models into TDNN and Conformer ASR systems for dysarthric and elderly speech recognition: a) input feature fusion between standard acoustic frontends and domain adapted wav2vec2.0 speech representations; b) frame-level joint decoding of TDNN systems separately trained using standard acoustic features alone and with additional wav2vec2.0 features; and c) multi-pass decoding involving the TDNN/Conformer system outputs to be rescored using domain adapted wav2vec2.0 models. In addition, domain adapted wav2vec2.0 representations are utilized in acoustic-to-articulatory (A2A) inversion to construct multi-modal dysarthric and elderly speech recognition systems. Experiments conducted on the UASpeech dysarthric and DementiaBank Pitt elderly speech corpora suggest TDNN and Conformer ASR systems integrated domain adapted wav2vec2.0 models consistently outperform the standalone wav2vec2.0 models by statistically significant WER reductions of 8.22% and 3.43% absolute (26.71% and 15.88% relative) on the two tasks respectively. The lowest published WERs of 22.56% (52.53% on very low intelligibility, 39.09% on unseen words) and 18.17% are obtained on the UASpeech test set of 16 dysarthric speakers, and the DementiaBank Pitt test set respectively.

Yi Wang, Jiajun Deng, Tianzi Wang et al.

Early diagnosis of Alzheimer's disease (AD) is crucial in facilitating preventive care and to delay further progression. Speech based automatic AD screening systems provide a non-intrusive and more scalable alternative to other clinical screening techniques. Textual embedding features produced by pre-trained language models (PLMs) such as BERT are widely used in such systems. However, PLM domain fine-tuning is commonly based on the masked word or sentence prediction costs that are inconsistent with the back-end AD detection task. To this end, this paper investigates the use of prompt-based fine-tuning of PLMs that consistently uses AD classification errors as the training objective function. Disfluency features based on hesitation or pause filler token frequencies are further incorporated into prompt phrases during PLM fine-tuning. The decision voting based combination among systems using different PLMs (BERT and RoBERTa) or systems with different fine-tuning paradigms (conventional masked-language modelling fine-tuning and prompt-based fine-tuning) is further applied. Mean, standard deviation and the maximum among accuracy scores over 15 experiment runs are adopted as performance measurements for the AD detection system. Mean detection accuracy of 84.20% (with std 2.09%, best 87.5%) and 82.64% (with std 4.0%, best 89.58%) were obtained using manual and ASR speech transcripts respectively on the ADReSS20 test set consisting of 48 elderly speakers.

Guinan Li, Jianwei Yu, Jiajun Deng et al.

Despite the rapid advance of automatic speech recognition (ASR) technologies, accurate recognition of cocktail party speech characterised by the interference from overlapping speakers, background noise and room reverberation remains a highly challenging task to date. Motivated by the invariance of visual modality to acoustic signal corruption, audio-visual speech enhancement techniques have been developed, although predominantly targeting overlapping speech separation and recognition tasks. In this paper, an audio-visual multi-channel speech separation, dereverberation and recognition approach featuring a full incorporation of visual information into all three stages of the system is proposed. The advantage of the additional visual modality over using audio only is demonstrated on two neural dereverberation approaches based on DNN-WPE and spectral mapping respectively. The learning cost function mismatch between the separation and dereverberation models and their integration with the back-end recognition system is minimised using fine-tuning on the MSE and LF-MMI criteria. Experiments conducted on the LRS2 dataset suggest that the proposed audio-visual multi-channel speech separation, dereverberation and recognition system outperforms the baseline audio-visual multi-channel speech separation and recognition system containing no dereverberation module by a statistically significant word error rate (WER) reduction of 2.06% absolute (8.77% relative).

Haonan Chen, Jingxiang Guo, Bangjun Wang et al.

Generalization remains a fundamental challenge in robotic manipulation. To tackle this challenge, recent Vision-Language-Action (VLA) models build policies on top of Vision-Language Models (VLMs), seeking to transfer their open-world semantic knowledge. However, their zero-shot capability lags significantly behind the base VLMs, as the instruction-vision-action data is too limited to cover diverse scenarios, tasks, and robot embodiments. In this work, we present Goal-VLA, a zero-shot framework that leverages Image-Generative VLMs as world models to generate desired goal states, from which the target object pose is derived to enable generalizable manipulation. The key insight is that object state representation is the golden interface, naturally separating a manipulation system into high-level and low-level policies. This representation abstracts away explicit action annotations, allowing the use of highly generalizable VLMs while simultaneously providing spatial cues for training-free low-level control. To further improve robustness, we introduce a Reflection-through-Synthesis process that iteratively validates and refines the generated goal image before execution. Both simulated and real-world experiments demonstrate that our \name achieves strong performance and inspiring generalizability in manipulation tasks. Supplementary materials are available at https://nus-lins-lab.github.io/goalvlaweb/.

Xiaomeng Chu, Jiajun Deng, Jianmin Ji et al.

The recent advance in multi-camera 3D object detection is featured by bird's-eye view (BEV) representation or object queries. However, the ill-posed transformation from image-plane view to 3D space inevitably causes feature clutter and distortion, making the objects blur into the background. To this end, we explore how to incorporate supplementary cues for differentiating objects in the transformed feature representation. Formally, we introduce OA-DET3D, a general plug-in module that improves 3D object detection by bringing object awareness into a variety of existing 3D object detection pipelines. Specifically, OA-DET3D boosts the representation of objects by leveraging object-centric depth information and foreground pseudo points. First, we use object-level supervision from the properties of each 3D bounding box to guide the network in learning the depth distribution. Next, we select foreground pixels using a 2D object detector and project them into 3D space for pseudo-voxel feature encoding. Finally, the object-aware depth features and pseudo-voxel features are incorporated into the BEV representation or query feature from the baseline model with a deformable attention mechanism. We conduct extensive experiments on the nuScenes dataset and Argoverse 2 dataset to validate the merits of OA-DET3D. Our method achieves consistent improvements over the BEV-based baselines in terms of both average precision and comprehensive detection score.

Zichen Yu, Changyong Shu, Jiajun Deng et al.

Given the capability of mitigating the long-tail deficiencies and intricate-shaped absence prevalent in 3D object detection, occupancy prediction has become a pivotal component in autonomous driving systems. However, the procession of three-dimensional voxel-level representations inevitably introduces large overhead in both memory and computation, obstructing the deployment of to-date occupancy prediction approaches. In contrast to the trend of making the model larger and more complicated, we argue that a desirable framework should be deployment-friendly to diverse chips while maintaining high precision. To this end, we propose a plug-and-play paradigm, namely FlashOCC, to consolidate rapid and memory-efficient occupancy prediction while maintaining high precision. Particularly, our FlashOCC makes two improvements based on the contemporary voxel-level occupancy prediction approaches. Firstly, the features are kept in the BEV, enabling the employment of efficient 2D convolutional layers for feature extraction. Secondly, a channel-to-height transformation is introduced to lift the output logits from the BEV into the 3D space. We apply the FlashOCC to diverse occupancy prediction baselines on the challenging Occ3D-nuScenes benchmarks and conduct extensive experiments to validate the effectiveness. The results substantiate the superiority of our plug-and-play paradigm over previous state-of-the-art methods in terms of precision, runtime efficiency, and memory costs, demonstrating its potential for deployment. The code will be made available.

Xiaomeng Chu, Jiajun Deng, Guoliang You et al.

We propose Radar-Camera fusion transformer (RaCFormer) to boost the accuracy of 3D object detection by the following insight. The Radar-Camera fusion in outdoor 3D scene perception is capped by the image-to-BEV transformation--if the depth of pixels is not accurately estimated, the naive combination of BEV features actually integrates unaligned visual content. To avoid this problem, we propose a query-based framework that enables adaptive sampling of instance-relevant features from both the bird's-eye view (BEV) and the original image view. Furthermore, we enhance system performance by two key designs: optimizing query initialization and strengthening the representational capacity of BEV. For the former, we introduce an adaptive circular distribution in polar coordinates to refine the initialization of object queries, allowing for a distance-based adjustment of query density. For the latter, we initially incorporate a radar-guided depth head to refine the transformation from image view to BEV. Subsequently, we focus on leveraging the Doppler effect of radar and introduce an implicit dynamic catcher to capture the temporal elements within the BEV. Extensive experiments on nuScenes and View-of-Delft (VoD) datasets validate the merits of our design. Remarkably, our method achieves superior results of 64.9% mAP and 70.2% NDS on nuScenes. RaCFormer also secures the state-of-the-art performance on the VoD dataset. Code is available at https://github.com/cxmomo/RaCFormer.

Hao Feng, Yuechen Wang, Wengang Zhou et al.

In this work, we propose a new framework, called Document Image Transformer (DocTr), to address the issue of geometry and illumination distortion of the document images. Specifically, DocTr consists of a geometric unwarping transformer and an illumination correction transformer. By setting a set of learned query embedding, the geometric unwarping transformer captures the global context of the document image by self-attention mechanism and decodes the pixel-wise displacement solution to correct the geometric distortion. After geometric unwarping, our illumination correction transformer further removes the shading artifacts to improve the visual quality and OCR accuracy. Extensive evaluations are conducted on several datasets, and superior results are reported against the state-of-the-art methods. Remarkably, our DocTr achieves 20.02% Character Error Rate (CER), a 15% absolute improvement over the state-of-the-art methods. Moreover, it also shows high efficiency on running time and parameter count. The results will be available at https://github.com/fh2019ustc/DocTr for further comparison.

Jiajun Deng, Wengang Zhou, Yanyong Zhang et al.

As an emerging data modal with precise distance sensing, LiDAR point clouds have been placed great expectations on 3D scene understanding. However, point clouds are always sparsely distributed in the 3D space, and with unstructured storage, which makes it difficult to represent them for effective 3D object detection. To this end, in this work, we regard point clouds as hollow-3D data and propose a new architecture, namely Hallucinated Hollow-3D R-CNN ($\text{H}^2$3D R-CNN), to address the problem of 3D object detection. In our approach, we first extract the multi-view features by sequentially projecting the point clouds into the perspective view and the bird-eye view. Then, we hallucinate the 3D representation by a novel bilaterally guided multi-view fusion block. Finally, the 3D objects are detected via a box refinement module with a novel Hierarchical Voxel RoI Pooling operation. The proposed $\text{H}^2$3D R-CNN provides a new angle to take full advantage of complementary information in the perspective view and the bird-eye view with an efficient framework. We evaluate our approach on the public KITTI Dataset and Waymo Open Dataset. Extensive experiments demonstrate the superiority of our method over the state-of-the-art algorithms with respect to both effectiveness and efficiency. The code will be made available at \url{https://github.com/djiajunustc/H-23D_R-CNN}.

Jiajun Deng, Zhengyuan Yang, Tianlang Chen et al.

In this paper, we present a neat yet effective transformer-based framework for visual grounding, namely TransVG, to address the task of grounding a language query to the corresponding region onto an image. The state-of-the-art methods, including two-stage or one-stage ones, rely on a complex module with manually-designed mechanisms to perform the query reasoning and multi-modal fusion. However, the involvement of certain mechanisms in fusion module design, such as query decomposition and image scene graph, makes the models easily overfit to datasets with specific scenarios, and limits the plenitudinous interaction between the visual-linguistic context. To avoid this caveat, we propose to establish the multi-modal correspondence by leveraging transformers, and empirically show that the complex fusion modules e.g., modular attention network, dynamic graph, and multi-modal tree) can be replaced by a simple stack of transformer encoder layers with higher performance. Moreover, we re-formulate the visual grounding as a direct coordinates regression problem and avoid making predictions out of a set of candidates i.e., region proposals or anchor boxes). Extensive experiments are conducted on five widely used datasets, and a series of state-of-the-art records are set by our TransVG. We build the benchmark of transformer-based visual grounding framework and make the code available at \url{https://github.com/djiajunustc/TransVG}.

Jiajun Deng, Shaoshuai Shi, Peiwei Li et al.

Recent advances on 3D object detection heavily rely on how the 3D data are represented, \emph{i.e.}, voxel-based or point-based representation. Many existing high performance 3D detectors are point-based because this structure can better retain precise point positions. Nevertheless, point-level features lead to high computation overheads due to unordered storage. In contrast, the voxel-based structure is better suited for feature extraction but often yields lower accuracy because the input data are divided into grids. In this paper, we take a slightly different viewpoint -- we find that precise positioning of raw points is not essential for high performance 3D object detection and that the coarse voxel granularity can also offer sufficient detection accuracy. Bearing this view in mind, we devise a simple but effective voxel-based framework, named Voxel R-CNN. By taking full advantage of voxel features in a two stage approach, our method achieves comparable detection accuracy with state-of-the-art point-based models, but at a fraction of the computation cost. Voxel R-CNN consists of a 3D backbone network, a 2D bird-eye-view (BEV) Region Proposal Network and a detect head. A voxel RoI pooling is devised to extract RoI features directly from voxel features for further refinement. Extensive experiments are conducted on the widely used KITTI Dataset and the more recent Waymo Open Dataset. Our results show that compared to existing voxel-based methods, Voxel R-CNN delivers a higher detection accuracy while maintaining a real-time frame processing rate, \emph{i.e}., at a speed of 25 FPS on an NVIDIA RTX 2080 Ti GPU. The code is available at \url{https://github.com/djiajunustc/Voxel-R-CNN}.

Jiajun Deng, Yingwei Pan, Ting Yao et al.

Single shot detectors that are potentially faster and simpler than two-stage detectors tend to be more applicable to object detection in videos. Nevertheless, the extension of such object detectors from image to video is not trivial especially when appearance deterioration exists in videos, \emph{e.g.}, motion blur or occlusion. A valid question is how to explore temporal coherence across frames for boosting detection. In this paper, we propose to address the problem by enhancing per-frame features through aggregation of neighboring frames. Specifically, we present Single Shot Video Object Detector (SSVD) -- a new architecture that novelly integrates feature aggregation into a one-stage detector for object detection in videos. Technically, SSVD takes Feature Pyramid Network (FPN) as backbone network to produce multi-scale features. Unlike the existing feature aggregation methods, SSVD, on one hand, estimates the motion and aggregates the nearby features along the motion path, and on the other, hallucinates features by directly sampling features from the adjacent frames in a two-stream structure. Extensive experiments are conducted on ImageNet VID dataset, and competitive results are reported when comparing to state-of-the-art approaches. More remarkably, for $448 \times 448$ input, SSVD achieves 79.2% mAP on ImageNet VID, by processing one frame in 85 ms on an Nvidia Titan X Pascal GPU. The code is available at \url{https://github.com/ddjiajun/SSVD}.

Tianlang Chen, Jiajun Deng, Jiebo Luo

Existing image-text matching approaches typically leverage triplet loss with online hard negatives to train the model. For each image or text anchor in a training mini-batch, the model is trained to distinguish between a positive and the most confusing negative of the anchor mined from the mini-batch (i.e. online hard negative). This strategy improves the model's capacity to discover fine-grained correspondences and non-correspondences between image and text inputs. However, the above approach has the following drawbacks: (1) the negative selection strategy still provides limited chances for the model to learn from very hard-to-distinguish cases. (2) The trained model has weak generalization capability from the training set to the testing set. (3) The penalty lacks hierarchy and adaptiveness for hard negatives with different "hardness" degrees. In this paper, we propose solutions by sampling negatives offline from the whole training set. It provides "harder" offline negatives than online hard negatives for the model to distinguish. Based on the offline hard negatives, a quintuplet loss is proposed to improve the model's generalization capability to distinguish positives and negatives. In addition, a novel loss function that combines the knowledge of positives, offline hard negatives and online hard negatives is created. It leverages offline hard negatives as the intermediary to adaptively penalize them based on their distance relations to the anchor. We evaluate the proposed training approach on three state-of-the-art image-text models on the MS-COCO and Flickr30K datasets. Significant performance improvements are observed for all the models, proving the effectiveness and generality of our approach. Code is available at https://github.com/sunnychencool/AOQ

Chunyat Wu, Jiajun Deng, Zhengxi Liu et al.

Although diffusion-based, non-autoregressive text-to-speech (TTS) systems have demonstrated impressive zero-shot synthesis capabilities, their efficacy is still hindered by two key challenges: the difficulty of text-speech alignment modeling and the high computational overhead of the iterative denoising process. To address these limitations, we propose ARCHI-TTS that features a dedicated semantic aligner to ensure robust temporal and semantic consistency between text and audio. To overcome high computational inference costs, ARCHI-TTS employs an efficient inference strategy that reuses encoder features across denoising steps, drastically accelerating synthesis without performance degradation. An auxiliary CTC loss applied to the condition encoder further enhances the semantic understanding. Experimental results demonstrate that ARCHI-TTS achieves a WER of 1.98% on LibriSpeech-PC test-clean, and 1.47%/1.42% on SeedTTS test-en/test-zh with a high inference efficiency, consistently outperforming recent state-of-the-art TTS systems.

Henan Wang, Hanxin Zhu, Tianyu He et al.

3D Gaussian Splatting (3DGS) has become an emerging technique with remarkable potential in 3D representation and image rendering. However, the substantial storage overhead of 3DGS significantly impedes its practical applications. In this work, we formulate the compact 3D Gaussian learning as an end-to-end Rate-Distortion Optimization (RDO) problem and propose RDO-Gaussian that can achieve flexible and continuous rate control. RDO-Gaussian addresses two main issues that exist in current schemes: 1) Different from prior endeavors that minimize the rate under the fixed distortion, we introduce dynamic pruning and entropy-constrained vector quantization (ECVQ) that optimize the rate and distortion at the same time. 2) Previous works treat the colors of each Gaussian equally, while we model the colors of different regions and materials with learnable numbers of parameters. We verify our method on both real and synthetic scenes, showcasing that RDO-Gaussian greatly reduces the size of 3D Gaussian over 40x, and surpasses existing methods in rate-distortion performance.

Jiajun Deng, Tianyu He, Li Jiang et al.

Current 3D Large Multimodal Models (3D LMMs) have shown tremendous potential in 3D-vision-based dialogue and reasoning. However, how to further enhance 3D LMMs to achieve fine-grained scene understanding and facilitate flexible human-agent interaction remains a challenging problem. In this work, we introduce 3D-LLaVA, a simple yet highly powerful 3D LMM designed to act as an intelligent assistant in comprehending, reasoning, and interacting with the 3D world. Unlike existing top-performing methods that rely on complicated pipelines-such as offline multi-view feature extraction or additional task-specific heads-3D-LLaVA adopts a minimalist design with integrated architecture and only takes point clouds as input. At the core of 3D-LLaVA is a new Omni Superpoint Transformer (OST), which integrates three functionalities: (1) a visual feature selector that converts and selects visual tokens, (2) a visual prompt encoder that embeds interactive visual prompts into the visual token space, and (3) a referring mask decoder that produces 3D masks based on text description. This versatile OST is empowered by the hybrid pretraining to obtain perception priors and leveraged as the visual connector that bridges the 3D data to the LLM. After performing unified instruction tuning, our 3D-LLaVA reports impressive results on various benchmarks.

Sha Zhang, Jiajun Deng, Lei Bai et al.

We present a hybrid-view-based knowledge distillation framework, termed HVDistill, to guide the feature learning of a point cloud neural network with a pre-trained image network in an unsupervised manner. By exploiting the geometric relationship between RGB cameras and LiDAR sensors, the correspondence between the two modalities based on both image-plane view and bird-eye view can be established, which facilitates representation learning. Specifically, the image-plane correspondences can be simply obtained by projecting the point clouds, while the bird-eye-view correspondences can be achieved by lifting pixels to the 3D space with the predicted depths under the supervision of projected point clouds. The image teacher networks provide rich semantics from the image-plane view and meanwhile acquire geometric information from the bird-eye view. Indeed, image features from the two views naturally complement each other and together can ameliorate the learned feature representation of the point cloud student networks. Moreover, with a self-supervised pre-trained 2D network, HVDistill requires neither 2D nor 3D annotations. We pre-train our model on nuScenes dataset and transfer it to several downstream tasks on nuScenes, SemanticKITTI, and KITTI datasets for evaluation. Extensive experimental results show that our method achieves consistent improvements over the baseline trained from scratch and significantly outperforms the existing schemes. Codes are available at git@github.com:zhangsha1024/HVDistill.git.

Yufei Yin, Lechao Cheng, Wengang Zhou et al.

In recent years, weakly supervised object detection (WSOD) has attracted much attention due to its low labeling cost. The success of recent WSOD models is often ascribed to the two-stage multi-class classification (MCC) task, i.e., multiple instance learning and online classification refinement. Despite achieving non-trivial progresses, these methods overlook potential classification ambiguities between these two MCC tasks and fail to leverage their unique strengths. In this work, we introduce a novel WSOD framework to ameliorate these two issues. For one thing, we propose a self-classification enhancement module that integrates intra-class binary classification (ICBC) to bridge the gap between the two distinct MCC tasks. The ICBC task enhances the network's discrimination between positive and mis-located samples in a class-wise manner and forges a mutually reinforcing relationship with the MCC task. For another, we propose a self-classification correction algorithm during inference, which combines the results of both MCC tasks to effectively reduce the mis-classified predictions. Extensive experiments on the prevalent VOC 2007 & 2012 datasets demonstrate the superior performance of our framework.

Shiqi Zhang, Sha Zhang, Jiajun Deng et al.

Existing open-vocabulary 3D semantic segmentation methods typically supervise 3D segmentation models by merging text-aligned features (e.g., CLIP) extracted from multi-view images onto 3D points. However, such approaches treat multi-view images merely as intermediaries for transferring open-vocabulary information, overlooking their rich semantic content and cross-view correspondences, which limits model effectiveness. To address this, we propose PGOV3D, a novel framework that introduces a Partial-to-Global curriculum for improving open-vocabulary 3D semantic segmentation. The key innovation lies in a two-stage training strategy. In the first stage, we pre-train the model on partial scenes that provide dense semantic information but relatively simple geometry. These partial point clouds are derived from multi-view RGB-D inputs via pixel-wise depth projection. To enable open-vocabulary learning, we leverage a multi-modal large language model (MLLM) and a 2D segmentation foundation model to generate open-vocabulary labels for each viewpoint, offering rich and aligned supervision. An auxiliary inter-frame consistency module is introduced to enforce feature consistency across varying viewpoints and enhance spatial understanding. In the second stage, we fine-tune the model on complete scene-level point clouds, which are sparser and structurally more complex. We aggregate the partial vocabularies associated with each scene and generate pseudo labels using the pre-trained model, effectively bridging the semantic gap between dense partial observations and large-scale 3D environments. Extensive experiments on ScanNet, ScanNet200, and S3DIS benchmarks demonstrate that PGOV3D achieves competitive performance in open-vocabulary 3D semantic segmentation.

Tianzi Wang, Xurong Xie, Zhaoqing Li et al.

This paper proposes a novel non-autoregressive (NAR) block-based Attention Mask Decoder (AMD) that flexibly balances performance-efficiency trade-offs for Conformer ASR systems. AMD performs parallel NAR inference within contiguous blocks of output labels that are concealed using attention masks, while conducting left-to-right AR prediction and history context amalgamation between blocks. A beam search algorithm is designed to leverage a dynamic fusion of CTC, AR Decoder, and AMD probabilities. Experiments on the LibriSpeech-100hr corpus suggest the tripartite Decoder incorporating the AMD module produces a maximum decoding speed-up ratio of 1.73x over the baseline CTC+AR decoding, while incurring no statistically significant word error rate (WER) increase on the test sets. When operating with the same decoding real time factors, statistically significant WER reductions of up to 0.7% and 0.3% absolute (5.3% and 6.1% relative) were obtained over the CTC+AR baseline.

Jiajun Deng, Sha Zhang, Feras Dayoub et al.

In this work, we present PoIFusion, a conceptually simple yet effective multi-modal 3D object detection framework to fuse the information of RGB images and LiDAR point clouds at the points of interest (PoIs). Different from the most accurate methods to date that transform multi-sensor data into a unified view or leverage the global attention mechanism to facilitate fusion, our approach maintains the view of each modality and obtains multi-modal features by computation-friendly projection and interpolation. In particular, our PoIFusion follows the paradigm of query-based object detection, formulating object queries as dynamic 3D boxes and generating a set of PoIs based on each query box. The PoIs serve as the keypoints to represent a 3D object and play the role of the basic units in multi-modal fusion. Specifically, we project PoIs into the view of each modality to sample the corresponding feature and integrate the multi-modal features at each PoI through a dynamic fusion block. Furthermore, the features of PoIs derived from the same query box are aggregated together to update the query feature. Our approach prevents information loss caused by view transformation and eliminates the computation-intensive global attention, making the multi-modal 3D object detector more applicable. We conducted extensive experiments on nuScenes and Argoverse2 datasets to evaluate our approach. Remarkably, the proposed approach achieves state-of-the-art results on both datasets without any bells and whistles, \emph{i.e.}, 74.9\% NDS and 73.4\% mAP on nuScenes, and 31.6\% CDS and 40.6\% mAP on Argoverse2. The code will be made available at \url{https://djiajunustc.github.io/projects/poifusion}.

Ning Wang, Jiajun Deng, Mingbo Jia

We present that visual grounding and image captioning, which perform as two mutually inverse processes, can be bridged together for collaborative training by careful designs. By consolidating this idea, we introduce CyCo, a cyclic-consistent learning framework to ameliorate the independent training pipelines of visual grounding and image captioning. The proposed framework (1) allows the semi-weakly supervised training of visual grounding; (2) improves the performance of fully supervised visual grounding; (3) yields a general captioning model that can describe arbitrary image regions. Extensive experiments show that our fully supervised grounding model achieves state-of-the-art performance, and the semi-weakly supervised one also exhibits competitive performance compared to the fully supervised counterparts. Our image captioning model has the capability to freely describe image regions and meanwhile shows impressive performance on prevalent captioning benchmarks.

Mengzhe Geng, Zengrui Jin, Tianzi Wang et al.

A key challenge in dysarthric speech recognition is the speaker-level diversity attributed to both speaker-identity associated factors such as gender, and speech impairment severity. Most prior researches on addressing this issue focused on using speaker-identity only. To this end, this paper proposes a novel set of techniques to use both severity and speaker-identity in dysarthric speech recognition: a) multitask training incorporating severity prediction error; b) speaker-severity aware auxiliary feature adaptation; and c) structured LHUC transforms separately conditioned on speaker-identity and severity. Experiments conducted on UASpeech suggest incorporating additional speech impairment severity into state-of-the-art hybrid DNN, E2E Conformer and pre-trained Wav2vec 2.0 ASR systems produced statistically significant WER reductions up to 4.78% (14.03% relative). Using the best system the lowest published WER of 17.82% (51.25% on very low intelligibility) was obtained on UASpeech.

Hanqi Zhu, Jiajun Deng, Yu Zhang et al.

It has been well recognized that fusing the complementary information from depth-aware LiDAR point clouds and semantic-rich stereo images would benefit 3D object detection. Nevertheless, it is not trivial to explore the inherently unnatural interaction between sparse 3D points and dense 2D pixels. To ease this difficulty, the recent proposals generally project the 3D points onto the 2D image plane to sample the image data and then aggregate the data at the points. However, this approach often suffers from the mismatch between the resolution of point clouds and RGB images, leading to sub-optimal performance. Specifically, taking the sparse points as the multi-modal data aggregation locations causes severe information loss for high-resolution images, which in turn undermines the effectiveness of multi-sensor fusion. In this paper, we present VPFNet -- a new architecture that cleverly aligns and aggregates the point cloud and image data at the `virtual' points. Particularly, with their density lying between that of the 3D points and 2D pixels, the virtual points can nicely bridge the resolution gap between the two sensors, and thus preserve more information for processing. Moreover, we also investigate the data augmentation techniques that can be applied to both point clouds and RGB images, as the data augmentation has made non-negligible contribution towards 3D object detectors to date. We have conducted extensive experiments on KITTI dataset, and have observed good performance compared to the state-of-the-art methods. Remarkably, our VPFNet achieves 83.21\% moderate 3D AP and 91.86\% moderate BEV AP on the KITTI test set, ranking the 1st since May 21th, 2021. The network design also takes computation efficiency into consideration -- we can achieve a FPS of 15 on a single NVIDIA RTX 2080Ti GPU. The code will be made available for reproduction and further investigation.

Hao Feng, Wengang Zhou, Jiajun Deng et al.

Compared with flatbed scanners, portable smartphones provide more convenience for physical document digitization. However, such digitized documents are often distorted due to uncontrolled physical deformations, camera positions, and illumination variations. To this end, we present DocScanner, a novel framework for document image rectification. Different from existing solutions, DocScanner addresses this issue by introducing a progressive learning mechanism. Specifically, DocScanner maintains a single estimate of the rectified image, which is progressively corrected with a recurrent architecture. The iterative refinements make DocScanner converge to a robust and superior rectification performance, while the lightweight recurrent architecture ensures the running efficiency. To further improve the rectification quality, based on the geometric priori between the distorted and the rectified images, a geometric regularization is introduced during training to further improve the performance. Extensive experiments are conducted on the Doc3D dataset and the DocUNet Benchmark dataset, and the quantitative and qualitative evaluation results verify the effectiveness of DocScanner, which outperforms previous methods on OCR accuracy, image similarity, and our proposed distortion metric by a considerable margin. Furthermore, our DocScanner shows superior efficiency in runtime latency and model size.

Xiaomeng Chu, Jiajun Deng, Yao Li et al.

As cameras are increasingly deployed in new application domains such as autonomous driving, performing 3D object detection on monocular images becomes an important task for visual scene understanding. Recent advances on monocular 3D object detection mainly rely on the ``pseudo-LiDAR'' generation, which performs monocular depth estimation and lifts the 2D pixels to pseudo 3D points. However, depth estimation from monocular images, due to its poor accuracy, leads to inevitable position shift of pseudo-LiDAR points within the object. Therefore, the predicted bounding boxes may suffer from inaccurate location and deformed shape. In this paper, we present a novel neighbor-voting method that incorporates neighbor predictions to ameliorate object detection from severely deformed pseudo-LiDAR point clouds. Specifically, each feature point around the object forms their own predictions, and then the ``consensus'' is achieved through voting. In this way, we can effectively combine the neighbors' predictions with local prediction and achieve more accurate 3D detection. To further enlarge the difference between the foreground region of interest (ROI) pseudo-LiDAR points and the background points, we also encode the ROI prediction scores of 2D foreground pixels into the corresponding pseudo-LiDAR points. We conduct extensive experiments on the KITTI benchmark to validate the merits of our proposed method. Our results on the bird's eye view detection outperform the state-of-the-art performance by a large margin, especially for the ``hard'' level detection.

Yuechen Wang, Jiajun Deng, Wengang Zhou et al.

Temporal language grounding (TLG) is a fundamental and challenging problem for vision and language understanding. Existing methods mainly focus on fully supervised setting with temporal boundary labels for training, which, however, suffers expensive cost of annotation. In this work, we are dedicated to weakly supervised TLG, where multiple description sentences are given to an untrimmed video without temporal boundary labels. In this task, it is critical to learn a strong cross-modal semantic alignment between sentence semantics and visual content. To this end, we introduce a novel weakly supervised temporal adjacent network (WSTAN) for temporal language grounding. Specifically, WSTAN learns cross-modal semantic alignment by exploiting temporal adjacent network in a multiple instance learning (MIL) paradigm, with a whole description paragraph as input. Moreover, we integrate a complementary branch into the framework, which explicitly refines the predictions with pseudo supervision from the MIL stage. An additional self-discriminating loss is devised on both the MIL branch and the complementary branch, aiming to enhance semantic discrimination by self-supervising. Extensive experiments are conducted on three widely used benchmark datasets, \emph{i.e.}, ActivityNet-Captions, Charades-STA, and DiDeMo, and the results demonstrate the effectiveness of our approach.

Shaoshuai Shi, Li Jiang, Jiajun Deng et al.

3D object detection is receiving increasing attention from both industry and academia thanks to its wide applications in various fields. In this paper, we propose Point-Voxel Region-based Convolution Neural Networks (PV-RCNNs) for 3D object detection on point clouds. First, we propose a novel 3D detector, PV-RCNN, which boosts the 3D detection performance by deeply integrating the feature learning of both point-based set abstraction and voxel-based sparse convolution through two novel steps, i.e., the voxel-to-keypoint scene encoding and the keypoint-to-grid RoI feature abstraction. Second, we propose an advanced framework, PV-RCNN++, for more efficient and accurate 3D object detection. It consists of two major improvements: sectorized proposal-centric sampling for efficiently producing more representative keypoints, and VectorPool aggregation for better aggregating local point features with much less resource consumption. With these two strategies, our PV-RCNN++ is about $3\times$ faster than PV-RCNN, while also achieving better performance. The experiments demonstrate that our proposed PV-RCNN++ framework achieves state-of-the-art 3D detection performance on the large-scale and highly-competitive Waymo Open Dataset with 10 FPS inference speed on the detection range of 150m * 150m.

Jiajun Deng, Yingwei Pan, Ting Yao et al.

It has been well recognized that modeling object-to-object relations would be helpful for object detection. Nevertheless, the problem is not trivial especially when exploring the interactions between objects to boost video object detectors. The difficulty originates from the aspect that reliable object relations in a video should depend on not only the objects in the present frame but also all the supportive objects extracted over a long range span of the video. In this paper, we introduce a new design to capture the interactions across the objects in spatio-temporal context. Specifically, we present Relation Distillation Networks (RDN) --- a new architecture that novelly aggregates and propagates object relation to augment object features for detection. Technically, object proposals are first generated via Region Proposal Networks (RPN). RDN then, on one hand, models object relation via multi-stage reasoning, and on the other, progressively distills relation through refining supportive object proposals with high objectness scores in a cascaded manner. The learnt relation verifies the efficacy on both improving object detection in each frame and box linking across frames. Extensive experiments are conducted on ImageNet VID dataset, and superior results are reported when comparing to state-of-the-art methods. More remarkably, our RDN achieves 81.8% and 83.2% mAP with ResNet-101 and ResNeXt-101, respectively. When further equipped with linking and rescoring, we obtain to-date the best reported mAP of 83.8% and 84.7%.