Hang Xu

Cuican Yu, Guansong Lu, Yihan Zeng et al. · tsinghua

Generating 3D faces from textual descriptions has a multitude of applications, such as gaming, movie, and robotics. Recent progresses have demonstrated the success of unconditional 3D face generation and text-to-3D shape generation. However, due to the limited text-3D face data pairs, text-driven 3D face generation remains an open problem. In this paper, we propose a text-guided 3D faces generation method, refer as TG-3DFace, for generating realistic 3D faces using text guidance. Specifically, we adopt an unconditional 3D face generation framework and equip it with text conditions, which learns the text-guided 3D face generation with only text-2D face data. On top of that, we propose two text-to-face cross-modal alignment techniques, including the global contrastive learning and the fine-grained alignment module, to facilitate high semantic consistency between generated 3D faces and input texts. Besides, we present directional classifier guidance during the inference process, which encourages creativity for out-of-domain generations. Compared to the existing methods, TG-3DFace creates more realistic and aesthetically pleasing 3D faces, boosting 9% multi-view consistency (MVIC) over Latent3D. The rendered face images generated by TG-3DFace achieve higher FID and CLIP score than text-to-2D face/image generation models, demonstrating our superiority in generating realistic and semantic-consistent textures.

Hang Xu, Chen Long, Wenxiao Zhang et al.

In this paper, we explore a novel framework, EGIInet (Explicitly Guided Information Interaction Network), a model for View-guided Point cloud Completion (ViPC) task, which aims to restore a complete point cloud from a partial one with a single view image. In comparison with previous methods that relied on the global semantics of input images, EGIInet efficiently combines the information from two modalities by leveraging the geometric nature of the completion task. Specifically, we propose an explicitly guided information interaction strategy supported by modal alignment for point cloud completion. First, in contrast to previous methods which simply use 2D and 3D backbones to encode features respectively, we unified the encoding process to promote modal alignment. Second, we propose a novel explicitly guided information interaction strategy that could help the network identify critical information within images, thus achieving better guidance for completion. Extensive experiments demonstrate the effectiveness of our framework, and we achieved a new state-of-the-art (+16% CD over XMFnet) in benchmark datasets despite using fewer parameters than the previous methods. The pre-trained model and code and are available at https://github.com/WHU-USI3DV/EGIInet.

Kaixin Cai, Pengzhen Ren, Yi Zhu et al.

Recently, semantic segmentation models trained with image-level text supervision have shown promising results in challenging open-world scenarios. However, these models still face difficulties in learning fine-grained semantic alignment at the pixel level and predicting accurate object masks. To address this issue, we propose MixReorg, a novel and straightforward pre-training paradigm for semantic segmentation that enhances a model's ability to reorganize patches mixed across images, exploring both local visual relevance and global semantic coherence. Our approach involves generating fine-grained patch-text pairs data by mixing image patches while preserving the correspondence between patches and text. The model is then trained to minimize the segmentation loss of the mixed images and the two contrastive losses of the original and restored features. With MixReorg as a mask learner, conventional text-supervised semantic segmentation models can achieve highly generalizable pixel-semantic alignment ability, which is crucial for open-world segmentation. After training with large-scale image-text data, MixReorg models can be applied directly to segment visual objects of arbitrary categories, without the need for further fine-tuning. Our proposed framework demonstrates strong performance on popular zero-shot semantic segmentation benchmarks, outperforming GroupViT by significant margins of 5.0%, 6.2%, 2.5%, and 3.4% mIoU on PASCAL VOC2012, PASCAL Context, MS COCO, and ADE20K, respectively.

Bingqian Lin, Yunshuang Nie, Ziming Wei et al.

Vision-and-Language Navigation (VLN), as a crucial research problem of Embodied AI, requires an embodied agent to navigate through complex 3D environments following natural language instructions. Recent research has highlighted the promising capacity of large language models (LLMs) in VLN by improving navigational reasoning accuracy and interpretability. However, their predominant use in an offline manner usually suffers from substantial domain gap between the VLN task and the LLM training corpus. This paper introduces a novel strategy called Navigational Chain-of-Thought (NavCoT), where we fulfill parameter-efficient in-domain training to enable self-guided navigational decision, leading to a significant mitigation of the domain gap in a cost-effective manner. Specifically, at each timestep, the LLM is prompted to forecast the navigational chain-of-thought by: 1) acting as a world model to imagine the next observation according to the instruction, 2) selecting the candidate observation that best aligns with the imagination, and 3) determining the action based on the reasoning from the prior steps. Through constructing formalized labels for training, the LLM can learn to generate desired and reasonable chain-of-thought outputs for improving the action decision. Experimental results across various training settings and popular VLN benchmarks (e.g., Room-to-Room (R2R), Room-across-Room (RxR), Room-for-Room (R4R)) show the significant superiority of NavCoT over the direct action prediction variants. Through simple parameter-efficient finetuning, our NavCoT outperforms a recent GPT4-based approach with ~7% relative improvement on the R2R dataset. We believe that NavCoT will help unlock more task-adaptive and scalable LLM-based embodied agents, which are helpful for developing real-world robotics applications. Code is available at https://github.com/expectorlin/NavCoT.

Hang Xu, Linjiang Huang, Feng Zhao

Test-time scaling (TTS) aims to achieve better results by increasing random sampling and evaluating samples based on rules and metrics. However, in text-to-image(T2I) diffusion models, most related works focus on search strategies and reward models, yet the impact of the stochastic characteristic of noise in T2I diffusion models on the method's performance remains unexplored. In this work, we analyze the effects of randomness in T2I diffusion models and explore a new format of randomness for TTS: text embedding perturbation, which couples with existing randomness like SDE-injected noise to enhance generative diversity and quality. We start with a frequency-domain analysis of these formats of randomness and their impact on generation, and find that these two randomness exhibit complementary behavior in the frequency domain: spatial noise favors low-frequency components (early steps), while text embedding perturbation enhances high-frequency details (later steps), thereby compensating for the potential limitations of spatial noise randomness in high-frequency manipulation. Concurrently, text embedding demonstrates varying levels of tolerance to perturbation across different dimensions of the generation process. Specifically, our method consists of two key designs: (1) Introducing step-based text embedding perturbation, combining frequency-guided noise schedules with spatial noise perturbation. (2) Adapting the perturbation intensity selectively based on their frequency-specific contributions to generation and tolerance to perturbation. Our approach can be seamlessly integrated into existing TTS methods and demonstrates significant improvements on multiple benchmarks with almost no additional computation. Code is available at \href{https://github.com/xuhang07/TEP-Diffusion}{https://github.com/xuhang07/TEP-Diffusion}.

Ziming Wei, Bingqian Lin, Yunshuang Nie et al.

Data scarcity is a long-standing challenge in the Vision-Language Navigation (VLN) field, which extremely hinders the generalization of agents to unseen environments. Previous works primarily rely on additional simulator data or web-collected images/videos to improve the generalization. However, the simulator environments still face limited diversity, and the web-collected data often requires extensive labor to remove the noise. In this paper, we propose a Rewriting-driven AugMentation (RAM) paradigm for VLN, which directly creates the unseen observation-instruction pairs via rewriting human-annotated training data. Benefiting from our rewriting mechanism, new observation-instruction pairs can be obtained in both simulator-free and labor-saving manners to promote generalization. Specifically, we first introduce Object-Enriched Observation Rewriting, where we combine Vision-Language Models (VLMs) and Large Language Models (LLMs) to derive rewritten object-enriched scene descriptions, enabling observation synthesis with diverse objects and spatial layouts via Text-to-Image Generation Models (T2IMs). Then, we propose Observation-Contrast Instruction Rewriting, which generates observation-aligned rewritten instructions by requiring LLMs to reason the difference between original and new observations. We further develop a mixing-then-focusing training strategy with a random observation cropping scheme, effectively enhancing data distribution diversity while suppressing augmentation data noise during training. Experiments on both the discrete environments (R2R, REVERIE, and R4R datasets) and continuous environments (R2R-CE dataset) show the superior performance and impressive generalization ability of our method. Code is available at https://github.com/SaDil13/VLN-RAM.

Kun Xiang, Zhili Liu, Terry Jingchen Zhang et al.

In this paper, we address the challenging task of multimodal mathematical reasoning by incorporating the notion of ``slow thinking'' into multimodal large language models (MLLMs). Our core idea is that models can learn to adaptively use different levels of reasoning to tackle questions of different complexity. We propose a novel paradigm of Self-structured Chain of Thought (SCoT), which comprises of minimal semantic atomic steps. Different from existing methods that rely on structured templates or free-form paradigms, our method can not only generate cognitive CoT structures for various complex tasks but also mitigates the phenomena of overthinking for easier tasks. To introduce structured reasoning into visual cognition, we further design a novel AtomThink framework with four key modules, including (i) a data engine to generate high-quality multimodal reasoning paths; (ii) a supervised fine-tuning (SFT) process with serialized inference data; (iii) a policy-guided multi-turn inference method; and (iv) an atomic capability metric to evaluate the single step utilization rate. We conduct extensive experiments to show that the proposed AtomThink significantly improves the performance of baseline MLLMs, achieving more than 10\% average accuracy gains on MathVista and MathVerse. Compared to state-of-the-art structured CoT approaches, our method not only achieves higher accuracy but also improves data utilization by 5 times and boosts inference efficiency by 85.3\%. Our code is now public available in https://github.com/Quinn777/AtomThink.

Chenhe Dong, Guangrun Wang, Hang Xu et al.

Pre-trained language models have shown remarkable results on various NLP tasks. Nevertheless, due to their bulky size and slow inference speed, it is hard to deploy them on edge devices. In this paper, we have a critical insight that improving the feed-forward network (FFN) in BERT has a higher gain than improving the multi-head attention (MHA) since the computational cost of FFN is 2$\sim$3 times larger than MHA. Hence, to compact BERT, we are devoted to designing efficient FFN as opposed to previous works that pay attention to MHA. Since FFN comprises a multilayer perceptron (MLP) that is essential in BERT optimization, we further design a thorough search space towards an advanced MLP and perform a coarse-to-fine mechanism to search for an efficient BERT architecture. Moreover, to accelerate searching and enhance model transferability, we employ a novel warm-up knowledge distillation strategy at each search stage. Extensive experiments show our searched EfficientBERT is 6.9$\times$ smaller and 4.4$\times$ faster than BERT$\rm_{BASE}$, and has competitive performances on GLUE and SQuAD Benchmarks. Concretely, EfficientBERT attains a 77.7 average score on GLUE \emph{test}, 0.7 higher than MobileBERT$\rm_{TINY}$, and achieves an 85.3/74.5 F1 score on SQuAD v1.1/v2.0 \emph{dev}, 3.2/2.7 higher than TinyBERT$_4$ even without data augmentation. The code is released at https://github.com/cheneydon/efficient-bert.

Xunlin Zhan, Yangxin Wu, Xiao Dong et al.

Nowadays, customer's demands for E-commerce are more diversified, which introduces more complications to the product retrieval industry. Previous methods are either subject to single-modal input or perform supervised image-level product retrieval, thus fail to accommodate real-life scenarios where enormous weakly annotated multi-modal data are present. In this paper, we investigate a more realistic setting that aims to perform weakly-supervised multi-modal instance-level product retrieval among fine-grained product categories. To promote the study of this challenging task, we contribute Product1M, one of the largest multi-modal cosmetic datasets for real-world instance-level retrieval. Notably, Product1M contains over 1 million image-caption pairs and consists of two sample types, i.e., single-product and multi-product samples, which encompass a wide variety of cosmetics brands. In addition to the great diversity, Product1M enjoys several appealing characteristics including fine-grained categories, complex combinations, and fuzzy correspondence that well mimic the real-world scenes. Moreover, we propose a novel model named Cross-modal contrAstive Product Transformer for instance-level prodUct REtrieval (CAPTURE), that excels in capturing the potential synergy between multi-modal inputs via a hybrid-stream transformer in a self-supervised manner.CAPTURE generates discriminative instance features via masked multi-modal learning as well as cross-modal contrastive pretraining and it outperforms several SOTA cross-modal baselines. Extensive ablation studies well demonstrate the effectiveness and the generalization capacity of our model. Dataset and codes are available at https: //github.com/zhanxlin/Product1M.

Peidong Liu, Gengwei Zhang, Bochao Wang et al.

Designing proper loss functions for vision tasks has been a long-standing research direction to advance the capability of existing models. For object detection, the well-established classification and regression loss functions have been carefully designed by considering diverse learning challenges. Inspired by the recent progress in network architecture search, it is interesting to explore the possibility of discovering new loss function formulations via directly searching the primitive operation combinations. So that the learned losses not only fit for diverse object detection challenges to alleviate huge human efforts, but also have better alignment with evaluation metric and good mathematical convergence property. Beyond the previous auto-loss works on face recognition and image classification, our work makes the first attempt to discover new loss functions for the challenging object detection from primitive operation levels. We propose an effective convergence-simulation driven evolutionary search algorithm, called CSE-Autoloss, for speeding up the search progress by regularizing the mathematical rationality of loss candidates via convergence property verification and model optimization simulation. CSE-Autoloss involves the search space that cover a wide range of the possible variants of existing losses and discovers best-searched loss function combination within a short time (around 1.5 wall-clock days). We conduct extensive evaluations of loss function search on popular detectors and validate the good generalization capability of searched losses across diverse architectures and datasets. Our experiments show that the best-discovered loss function combinations outperform default combinations by 1.1% and 0.8% in terms of mAP for two-stage and one-stage detectors on COCO respectively. Our searched losses are available at https://github.com/PerdonLiu/CSE-Autoloss.

Bochao Wang, Hang Xu, Jiajin Zhang et al.

Automated Machine Learning (AutoML) is an important industrial solution for automatic discovery and deployment of the machine learning models. However, designing an integrated AutoML system faces four great challenges of configurability, scalability, integrability, and platform diversity. In this work, we present VEGA, an efficient and comprehensive AutoML framework that is compatible and optimized for multiple hardware platforms. a) The VEGA pipeline integrates various modules of AutoML, including Neural Architecture Search (NAS), Hyperparameter Optimization (HPO), Auto Data Augmentation, Model Compression, and Fully Train. b) To support a variety of search algorithms and tasks, we design a novel fine-grained search space and its description language to enable easy adaptation to different search algorithms and tasks. c) We abstract the common components of deep learning frameworks into a unified interface. VEGA can be executed with multiple back-ends and hardwares. Extensive benchmark experiments on multiple tasks demonstrate that VEGA can improve the existing AutoML algorithms and discover new high-performance models against SOTA methods, e.g. the searched DNet model zoo for Ascend 10x faster than EfficientNet-B5 and 9.2x faster than RegNetX-32GF on ImageNet. VEGA is open-sourced at https://github.com/huawei-noah/vega.

Yangxin Wu, Gengwei Zhang, Hang Xu et al.

Panoptic segmentation is posed as a new popular test-bed for the state-of-the-art holistic scene understanding methods with the requirement of simultaneously segmenting both foreground things and background stuff. The state-of-the-art panoptic segmentation network exhibits high structural complexity in different network components, i.e. backbone, proposal-based foreground branch, segmentation-based background branch, and feature fusion module across branches, which heavily relies on expert knowledge and tedious trials. In this work, we propose an efficient, cooperative and highly automated framework to simultaneously search for all main components including backbone, segmentation branches, and feature fusion module in a unified panoptic segmentation pipeline based on the prevailing one-shot Network Architecture Search (NAS) paradigm. Notably, we extend the common single-task NAS into the multi-component scenario by taking the advantage of the newly proposed intra-modular search space and problem-oriented inter-modular search space, which helps us to obtain an optimal network architecture that not only performs well in both instance segmentation and semantic segmentation tasks but also be aware of the reciprocal relations between foreground things and background stuff classes. To relieve the vast computation burden incurred by applying NAS to complicated network architectures, we present a novel path-priority greedy search policy to find a robust, transferrable architecture with significantly reduced searching overhead. Our searched architecture, namely Auto-Panoptic, achieves the new state-of-the-art on the challenging COCO and ADE20K benchmarks. Moreover, extensive experiments are conducted to demonstrate the effectiveness of path-priority policy and transferability of Auto-Panoptic across different datasets. Codes and models are available at: https://github.com/Jacobew/AutoPanoptic.

Hang Xu, Shaoju Wang, Xinyue Cai et al.

We address the curve lane detection problem which poses more realistic challenges than conventional lane detection for better facilitating modern assisted/autonomous driving systems. Current hand-designed lane detection methods are not robust enough to capture the curve lanes especially the remote parts due to the lack of modeling both long-range contextual information and detailed curve trajectory. In this paper, we propose a novel lane-sensitive architecture search framework named CurveLane-NAS to automatically capture both long-ranged coherent and accurate short-range curve information while unifying both architecture search and post-processing on curve lane predictions via point blending. It consists of three search modules: a) a feature fusion search module to find a better fusion of the local and global context for multi-level hierarchy features; b) an elastic backbone search module to explore an efficient feature extractor with good semantics and latency; c) an adaptive point blending module to search a multi-level post-processing refinement strategy to combine multi-scale head prediction. The unified framework ensures lane-sensitive predictions by the mutual guidance between NAS and adaptive point blending. Furthermore, we also steer forward to release a more challenging benchmark named CurveLanes for addressing the most difficult curve lanes. It consists of 150K images with 680K labels.The new dataset can be downloaded at github.com/xbjxh/CurveLanes (already anonymized for this submission). Experiments on the new CurveLanes show that the SOTA lane detection methods suffer substantial performance drop while our model can still reach an 80+% F1-score. Extensive experiments on traditional lane benchmarks such as CULane also demonstrate the superiority of our CurveLane-NAS, e.g. achieving a new SOTA 74.8% F1-score on CULane.

Chenhan Jiang, Hang Xu, Xiangdan Liang et al.

The dominant object detection approaches treat the recognition of each region separately and overlook crucial semantic correlations between objects in one scene. This paradigm leads to substantial performance drop when facing heavy long-tail problems, where very few samples are available for rare classes and plenty of confusing categories exists. We exploit diverse human commonsense knowledge for reasoning over large-scale object categories and reaching semantic coherency within one image. Particularly, we present Hybrid Knowledge Routed Modules (HKRM) that incorporates the reasoning routed by two kinds of knowledge forms: an explicit knowledge module for structured constraints that are summarized with linguistic knowledge (e.g. shared attributes, relationships) about concepts; and an implicit knowledge module that depicts some implicit constraints (e.g. common spatial layouts). By functioning over a region-to-region graph, both modules can be individualized and adapted to coordinate with visual patterns in each image, guided by specific knowledge forms. HKRM are light-weight, general-purpose and extensible by easily incorporating multiple knowledge to endow any detection networks the ability of global semantic reasoning. Experiments on large-scale object detection benchmarks show HKRM obtains around 34.5% improvement on VisualGenome (1000 categories) and 30.4% on ADE in terms of mAP. Codes and trained model can be found in https://github.com/chanyn/HKRM.

Youpeng Wen, Junfan Lin, Yi Zhu et al.

Recent advancements utilizing large-scale video data for learning video generation models demonstrate significant potential in understanding complex physical dynamics. It suggests the feasibility of leveraging diverse robot trajectory data to develop a unified, dynamics-aware model to enhance robot manipulation. However, given the relatively small amount of available robot data, directly fitting data without considering the relationship between visual observations and actions could lead to suboptimal data utilization. To this end, we propose VidMan (Video Diffusion for Robot Manipulation), a novel framework that employs a two-stage training mechanism inspired by dual-process theory from neuroscience to enhance stability and improve data utilization efficiency. Specifically, in the first stage, VidMan is pre-trained on the Open X-Embodiment dataset (OXE) for predicting future visual trajectories in a video denoising diffusion manner, enabling the model to develop a long horizontal awareness of the environment's dynamics. In the second stage, a flexible yet effective layer-wise self-attention adapter is introduced to transform VidMan into an efficient inverse dynamics model that predicts action modulated by the implicit dynamics knowledge via parameter sharing. Our VidMan framework outperforms state-of-the-art baseline model GR-1 on the CALVIN benchmark, achieving a 11.7% relative improvement, and demonstrates over 9% precision gains on the OXE small-scale dataset. These results provide compelling evidence that world models can significantly enhance the precision of robot action prediction. Codes and models will be public.

Xinpeng Ding, Kui Zhang, Jianhua Han et al.

Direct Preference Optimization (DPO) helps reduce hallucinations in Video Multimodal Large Language Models (VLLMs), but its reliance on offline preference data limits adaptability and fails to capture true video-response misalignment. We propose Video Direct Preference Optimization (VDPO), an online preference learning framework that eliminates the need for preference annotation by leveraging video augmentations to generate rejected samples while keeping responses fixed. However, selecting effective augmentations is non-trivial, as some clips may be semantically identical to the original under specific prompts, leading to false rejections and disrupting alignment. To address this, we introduce Prompt-aware Multi-instance Learning VDPO (PaMi-VDPO), which selects augmentations based on prompt context. Instead of a single rejection, we construct a candidate set of augmented clips and apply a close-to-far selection strategy, initially ensuring all clips are semantically relevant while then prioritizing the most prompt-aware distinct clip. This allows the model to better capture meaningful visual differences, mitigating hallucinations, while avoiding false rejections, and improving alignment. PaMi-VDPOseamlessly integrates into existing VLLMs without additional parameters, GPT-4/human supervision. With only 10k SFT data, it improves the base model by 5.3% on VideoHallucer, surpassing GPT-4o, while maintaining stable performance on general video benchmarks.

Lewei Yao, Runhui Huang, Lu Hou et al.

Unsupervised large-scale vision-language pre-training has shown promising advances on various downstream tasks. Existing methods often model the cross-modal interaction either via the similarity of the global feature of each modality which misses sufficient information, or finer-grained interactions using cross/self-attention upon visual and textual tokens. However, cross/self-attention suffers from inferior efficiency in both training and inference. In this paper, we introduce a large-scale Fine-grained Interactive Language-Image Pre-training (FILIP) to achieve finer-level alignment through a cross-modal late interaction mechanism, which uses a token-wise maximum similarity between visual and textual tokens to guide the contrastive objective. FILIP successfully leverages the finer-grained expressiveness between image patches and textual words by modifying only contrastive loss, while simultaneously gaining the ability to pre-compute image and text representations offline at inference, keeping both large-scale training and inference efficient. Furthermore, we construct a new large-scale image-text pair dataset called FILIP300M for pre-training. Experiments show that FILIP achieves state-of-the-art performance on multiple downstream vision-language tasks including zero-shot image classification and image-text retrieval. The visualization on word-patch alignment further shows that FILIP can learn meaningful fine-grained features with promising localization ability.

Jiageng Mao, Minzhe Niu, Haoyue Bai et al.

We present a flexible and high-performance framework, named Pyramid R-CNN, for two-stage 3D object detection from point clouds. Current approaches generally rely on the points or voxels of interest for RoI feature extraction on the second stage, but cannot effectively handle the sparsity and non-uniform distribution of those points, and this may result in failures in detecting objects that are far away. To resolve the problems, we propose a novel second-stage module, named pyramid RoI head, to adaptively learn the features from the sparse points of interest. The pyramid RoI head consists of three key components. Firstly, we propose the RoI-grid Pyramid, which mitigates the sparsity problem by extensively collecting points of interest for each RoI in a pyramid manner. Secondly, we propose RoI-grid Attention, a new operation that can encode richer information from sparse points by incorporating conventional attention-based and graph-based point operators into a unified formulation. Thirdly, we propose the Density-Aware Radius Prediction (DARP) module, which can adapt to different point density levels by dynamically adjusting the focusing range of RoIs. Combining the three components, our pyramid RoI head is robust to the sparse and imbalanced circumstances, and can be applied upon various 3D backbones to consistently boost the detection performance. Extensive experiments show that Pyramid R-CNN outperforms the state-of-the-art 3D detection models by a large margin on both the KITTI dataset and the Waymo Open dataset.

Jiageng Mao, Yujing Xue, Minzhe Niu et al.

We present Voxel Transformer (VoTr), a novel and effective voxel-based Transformer backbone for 3D object detection from point clouds. Conventional 3D convolutional backbones in voxel-based 3D detectors cannot efficiently capture large context information, which is crucial for object recognition and localization, owing to the limited receptive fields. In this paper, we resolve the problem by introducing a Transformer-based architecture that enables long-range relationships between voxels by self-attention. Given the fact that non-empty voxels are naturally sparse but numerous, directly applying standard Transformer on voxels is non-trivial. To this end, we propose the sparse voxel module and the submanifold voxel module, which can operate on the empty and non-empty voxel positions effectively. To further enlarge the attention range while maintaining comparable computational overhead to the convolutional counterparts, we propose two attention mechanisms for multi-head attention in those two modules: Local Attention and Dilated Attention, and we further propose Fast Voxel Query to accelerate the querying process in multi-head attention. VoTr contains a series of sparse and submanifold voxel modules and can be applied in most voxel-based detectors. Our proposed VoTr shows consistent improvement over the convolutional baselines while maintaining computational efficiency on the KITTI dataset and the Waymo Open dataset.

Muhammad Awais, Fengwei Zhou, Hang Xu et al.

Extensive Unsupervised Domain Adaptation (UDA) studies have shown great success in practice by learning transferable representations across a labeled source domain and an unlabeled target domain with deep models. However, previous works focus on improving the generalization ability of UDA models on clean examples without considering the adversarial robustness, which is crucial in real-world applications. Conventional adversarial training methods are not suitable for the adversarial robustness on the unlabeled target domain of UDA since they train models with adversarial examples generated by the supervised loss function. In this work, we leverage intermediate representations learned by multiple robust ImageNet models to improve the robustness of UDA models. Our method works by aligning the features of the UDA model with the robust features learned by ImageNet pre-trained models along with domain adaptation training. It utilizes both labeled and unlabeled domains and instills robustness without any adversarial intervention or label requirement during domain adaptation training. Experimental results show that our method significantly improves adversarial robustness compared to the baseline while keeping clean accuracy on various UDA benchmarks.

Lewei Yao, Renjie Pi, Hang Xu et al.

In this paper, we investigate the knowledge distillation (KD) strategy for object detection and propose an effective framework applicable to both homogeneous and heterogeneous student-teacher pairs. The conventional feature imitation paradigm introduces imitation masks to focus on informative foreground areas while excluding the background noises. However, we find that those methods fail to fully utilize the semantic information in all feature pyramid levels, which leads to inefficiency for knowledge distillation between FPN-based detectors. To this end, we propose a novel semantic-guided feature imitation technique, which automatically performs soft matching between feature pairs across all pyramid levels to provide the optimal guidance to the student. To push the envelop even further, we introduce contrastive distillation to effectively capture the information encoded in the relationship between different feature regions. Finally, we propose a generalized detection KD pipeline, which is capable of distilling both homogeneous and heterogeneous detector pairs. Our method consistently outperforms the existing detection KD techniques, and works when (1) components in the framework are used separately and in conjunction; (2) for both homogeneous and heterogenous student-teacher pairs and (3) on multiple detection benchmarks. With a powerful X101-FasterRCNN-Instaboost detector as the teacher, R50-FasterRCNN reaches 44.0% AP, R50-RetinaNet reaches 43.3% AP and R50-FCOS reaches 43.1% AP on COCO dataset.

Hang Xu, Ning Kang, Gengwei Zhang et al.

Fine-tuning from pre-trained ImageNet models has been a simple, effective, and popular approach for various computer vision tasks. The common practice of fine-tuning is to adopt a default hyperparameter setting with a fixed pre-trained model, while both of them are not optimized for specific tasks and time constraints. Moreover, in cloud computing or GPU clusters where the tasks arrive sequentially in a stream, faster online fine-tuning is a more desired and realistic strategy for saving money, energy consumption, and CO2 emission. In this paper, we propose a joint Neural Architecture Search and Online Adaption framework named NASOA towards a faster task-oriented fine-tuning upon the request of users. Specifically, NASOA first adopts an offline NAS to identify a group of training-efficient networks to form a pretrained model zoo. We propose a novel joint block and macro-level search space to enable a flexible and efficient search. Then, by estimating fine-tuning performance via an adaptive model by accumulating experience from the past tasks, an online schedule generator is proposed to pick up the most suitable model and generate a personalized training regime with respect to each desired task in a one-shot fashion. The resulting model zoo is more training efficient than SOTA models, e.g. 6x faster than RegNetY-16GF, and 1.7x faster than EfficientNetB3. Experiments on multiple datasets also show that NASOA achieves much better fine-tuning results, i.e. improving around 2.1% accuracy than the best performance in RegNet series under various constraints and tasks; 40x faster compared to the BOHB.

Jianhua Han, Xiwen Liang, Hang Xu et al.

Aiming at facilitating a real-world, ever-evolving and scalable autonomous driving system, we present a large-scale dataset for standardizing the evaluation of different self-supervised and semi-supervised approaches by learning from raw data, which is the first and largest dataset to date. Existing autonomous driving systems heavily rely on `perfect' visual perception models (i.e., detection) trained using extensive annotated data to ensure safety. However, it is unrealistic to elaborately label instances of all scenarios and circumstances (i.e., night, extreme weather, cities) when deploying a robust autonomous driving system. Motivated by recent advances of self-supervised and semi-supervised learning, a promising direction is to learn a robust detection model by collaboratively exploiting large-scale unlabeled data and few labeled data. Existing datasets either provide only a small amount of data or covers limited domains with full annotation, hindering the exploration of large-scale pre-trained models. Here, we release a Large-Scale 2D Self/semi-supervised Object Detection dataset for Autonomous driving, named as SODA10M, containing 10 million unlabeled images and 20K images labeled with 6 representative object categories. To improve diversity, the images are collected within 27833 driving hours under different weather conditions, periods and location scenes of 32 different cities. We provide extensive experiments and deep analyses of existing popular self/semi-supervised approaches, and give some interesting findings in autonomous driving scope. Experiments show that SODA10M can serve as a promising pre-training dataset for different self-supervised learning methods, which gives superior performance when fine-tuning with different downstream tasks (i.e., detection, semantic/instance segmentation) in autonomous driving domain. More information can refer to https://soda-2d.github.io.

Jiageng Mao, Minzhe Niu, Chenhan Jiang et al.

Current perception models in autonomous driving have become notorious for greatly relying on a mass of annotated data to cover unseen cases and address the long-tail problem. On the other hand, learning from unlabeled large-scale collected data and incrementally self-training powerful recognition models have received increasing attention and may become the solutions of next-generation industry-level powerful and robust perception models in autonomous driving. However, the research community generally suffered from data inadequacy of those essential real-world scene data, which hampers the future exploration of fully/semi/self-supervised methods for 3D perception. In this paper, we introduce the ONCE (One millioN sCenEs) dataset for 3D object detection in the autonomous driving scenario. The ONCE dataset consists of 1 million LiDAR scenes and 7 million corresponding camera images. The data is selected from 144 driving hours, which is 20x longer than the largest 3D autonomous driving dataset available (e.g. nuScenes and Waymo), and it is collected across a range of different areas, periods and weather conditions. To facilitate future research on exploiting unlabeled data for 3D detection, we additionally provide a benchmark in which we reproduce and evaluate a variety of self-supervised and semi-supervised methods on the ONCE dataset. We conduct extensive analyses on those methods and provide valuable observations on their performance related to the scale of used data. Data, code, and more information are available at https://once-for-auto-driving.github.io/index.html.

Lewei Yao, Renjie Pi, Hang Xu et al.

We propose Joint-DetNAS, a unified NAS framework for object detection, which integrates 3 key components: Neural Architecture Search, pruning, and Knowledge Distillation. Instead of naively pipelining these techniques, our Joint-DetNAS optimizes them jointly. The algorithm consists of two core processes: student morphism optimizes the student's architecture and removes the redundant parameters, while dynamic distillation aims to find the optimal matching teacher. For student morphism, weight inheritance strategy is adopted, allowing the student to flexibly update its architecture while fully utilize the predecessor's weights, which considerably accelerates the search; To facilitate dynamic distillation, an elastic teacher pool is trained via integrated progressive shrinking strategy, from which teacher detectors can be sampled without additional cost in subsequent searches. Given a base detector as the input, our algorithm directly outputs the derived student detector with high performance without additional training. Experiments demonstrate that our Joint-DetNAS outperforms the naive pipelining approach by a great margin. Given a classic R101-FPN as the base detector, Joint-DetNAS is able to boost its mAP from 41.4 to 43.9 on MS COCO and reduce the latency by 47%, which is on par with the SOTA EfficientDet while requiring less search cost. We hope our proposed method can provide the community with a new way of jointly optimizing NAS, KD and pruning.

Yawen Duan, Xin Chen, Hang Xu et al.

Recent breakthroughs of Neural Architecture Search (NAS) extend the field's research scope towards a broader range of vision tasks and more diversified search spaces. While existing NAS methods mostly design architectures on a single task, algorithms that look beyond single-task search are surging to pursue a more efficient and universal solution across various tasks. Many of them leverage transfer learning and seek to preserve, reuse, and refine network design knowledge to achieve higher efficiency in future tasks. However, the enormous computational cost and experiment complexity of cross-task NAS are imposing barriers for valuable research in this direction. Existing NAS benchmarks all focus on one type of vision task, i.e., classification. In this work, we propose TransNAS-Bench-101, a benchmark dataset containing network performance across seven tasks, covering classification, regression, pixel-level prediction, and self-supervised tasks. This diversity provides opportunities to transfer NAS methods among tasks and allows for more complex transfer schemes to evolve. We explore two fundamentally different types of search space: cell-level search space and macro-level search space. With 7,352 backbones evaluated on seven tasks, 51,464 trained models with detailed training information are provided. With TransNAS-Bench-101, we hope to encourage the advent of exceptional NAS algorithms that raise cross-task search efficiency and generalizability to the next level. Our dataset file will be available at Mindspore, VEGA.

Wenqi Shao, Hang Yu, Zhaoyang Zhang et al.

This work presents a probabilistic channel pruning method to accelerate Convolutional Neural Networks (CNNs). Previous pruning methods often zero out unimportant channels in training in a deterministic manner, which reduces CNN's learning capacity and results in suboptimal performance. To address this problem, we develop a probability-based pruning algorithm, called batch whitening channel pruning (BWCP), which can stochastically discard unimportant channels by modeling the probability of a channel being activated. BWCP has several merits. (1) It simultaneously trains and prunes CNNs from scratch in a probabilistic way, exploring larger network space than deterministic methods. (2) BWCP is empowered by the proposed batch whitening tool, which is able to empirically and theoretically increase the activation probability of useful channels while keeping unimportant channels unchanged without adding any extra parameters and computational cost in inference. (3) Extensive experiments on CIFAR-10, CIFAR-100, and ImageNet with various network architectures show that BWCP outperforms its counterparts by achieving better accuracy given limited computational budgets. For example, ResNet50 pruned by BWCP has only 0.70\% Top-1 accuracy drop on ImageNet, while reducing 43.1\% FLOPs of the plain ResNet50.

Xiao Zhou, Weizhong Zhang, Hang Xu et al.

Weight pruning is an effective technique to reduce the model size and inference time for deep neural networks in real-world deployments. However, since magnitudes and relative importance of weights are very different for different layers of a neural network, existing methods rely on either manual tuning or handcrafted heuristic rules to find appropriate pruning rates individually for each layer. This approach generally leads to suboptimal performance. In this paper, by directly working on the probability space, we propose an effective network sparsification method called {\it probabilistic masking} (ProbMask), which solves a natural sparsification formulation under global sparsity constraint. The key idea is to use probability as a global criterion for all layers to measure the weight importance. An appealing feature of ProbMask is that the amounts of weight redundancy can be learned automatically via our constraint and thus we avoid the problem of tuning pruning rates individually for different layers in a network. Extensive experimental results on CIFAR-10/100 and ImageNet demonstrate that our method is highly effective, and can outperform previous state-of-the-art methods by a significant margin, especially in the high pruning rate situation. Notably, the gap of Top-1 accuracy between our ProbMask and existing methods can be up to 10\%. As a by-product, we show ProbMask is also highly effective in identifying supermasks, which are subnetworks with high performance in a randomly weighted dense neural network.

Han Shi, Jiahui Gao, Xiaozhe Ren et al.

Transformer-based models are popularly used in natural language processing (NLP). Its core component, self-attention, has aroused widespread interest. To understand the self-attention mechanism, a direct method is to visualize the attention map of a pre-trained model. Based on the patterns observed, a series of efficient Transformers with different sparse attention masks have been proposed. From a theoretical perspective, universal approximability of Transformer-based models is also recently proved. However, the above understanding and analysis of self-attention is based on a pre-trained model. To rethink the importance analysis in self-attention, we study the significance of different positions in attention matrix during pre-training. A surprising result is that diagonal elements in the attention map are the least important compared with other attention positions. We provide a proof showing that these diagonal elements can indeed be removed without deteriorating model performance. Furthermore, we propose a Differentiable Attention Mask (DAM) algorithm, which further guides the design of the SparseBERT. Extensive experiments verify our interesting findings and illustrate the effect of the proposed algorithm.

Gengwei Zhang, Yiming Gao, Hang Xu et al.

Panoptic segmentation that unifies instance segmentation and semantic segmentation has recently attracted increasing attention. While most existing methods focus on designing novel architectures, we steer toward a different perspective: performing automated multi-loss adaptation (named Ada-Segment) on the fly to flexibly adjust multiple training losses over the course of training using a controller trained to capture the learning dynamics. This offers a few advantages: it bypasses manual tuning of the sensitive loss combination, a decisive factor for panoptic segmentation; it allows to explicitly model the learning dynamics, and reconcile the learning of multiple objectives (up to ten in our experiments); with an end-to-end architecture, it generalizes to different datasets without the need of re-tuning hyperparameters or re-adjusting the training process laboriously. Our Ada-Segment brings 2.7% panoptic quality (PQ) improvement on COCO val split from the vanilla baseline, achieving the state-of-the-art 48.5% PQ on COCO test-dev split and 32.9% PQ on ADE20K dataset. The extensive ablation studies reveal the ever-changing dynamics throughout the training process, necessitating the incorporation of an automated and adaptive learning strategy as presented in this paper.

Okan Köpüklü, Jiapeng Zheng, Hang Xu et al.

Distracted drivers are more likely to fail to anticipate hazards, which result in car accidents. Therefore, detecting anomalies in drivers' actions (i.e., any action deviating from normal driving) contains the utmost importance to reduce driver-related accidents. However, there are unbounded many anomalous actions that a driver can do while driving, which leads to an 'open set recognition' problem. Accordingly, instead of recognizing a set of anomalous actions that are commonly defined by previous dataset providers, in this work, we propose a contrastive learning approach to learn a metric to differentiate normal driving from anomalous driving. For this task, we introduce a new video-based benchmark, the Driver Anomaly Detection (DAD) dataset, which contains normal driving videos together with a set of anomalous actions in its training set. In the test set of the DAD dataset, there are unseen anomalous actions that still need to be winnowed out from normal driving. Our method reaches 0.9673 AUC on the test set, demonstrating the effectiveness of the contrastive learning approach on the anomaly detection task. Our dataset, codes and pre-trained models are publicly available.

Xin Chen, Yawen Duan, Zewei Chen et al.

Neural Architecture Search (NAS) achieved many breakthroughs in recent years. In spite of its remarkable progress, many algorithms are restricted to particular search spaces. They also lack efficient mechanisms to reuse knowledge when confronting multiple tasks. These challenges preclude their applicability, and motivate our proposal of CATCH, a novel Context-bAsed meTa reinforcement learning (RL) algorithm for transferrable arChitecture searcH. The combination of meta-learning and RL allows CATCH to efficiently adapt to new tasks while being agnostic to search spaces. CATCH utilizes a probabilistic encoder to encode task properties into latent context variables, which then guide CATCH's controller to quickly "catch" top-performing networks. The contexts also assist a network evaluator in filtering inferior candidates and speed up learning. Extensive experiments demonstrate CATCH's universality and search efficiency over many other widely-recognized algorithms. It is also capable of handling cross-domain architecture search as competitive networks on ImageNet, COCO, and Cityscapes are identified. This is the first work to our knowledge that proposes an efficient transferrable NAS solution while maintaining robustness across various settings.

Wenshuo Ma, Tingzhong Tian, Hang Xu et al.

Most state-of-the-art object detection systems follow an anchor-based diagram. Anchor boxes are densely proposed over the images and the network is trained to predict the boxes position offset as well as the classification confidence. Existing systems pre-define anchor box shapes and sizes and ad-hoc heuristic adjustments are used to define the anchor configurations. However, this might be sub-optimal or even wrong when a new dataset or a new model is adopted. In this paper, we study the problem of automatically optimizing anchor boxes for object detection. We first demonstrate that the number of anchors, anchor scales and ratios are crucial factors for a reliable object detection system. By carefully analyzing the existing bounding box patterns on the feature hierarchy, we design a flexible and tight hyper-parameter space for anchor configurations. Then we propose a novel hyper-parameter optimization method named AABO to determine more appropriate anchor boxes for a certain dataset, in which Bayesian Optimization and subsampling method are combined to achieve precise and efficient anchor configuration optimization. Experiments demonstrate the effectiveness of our proposed method on different detectors and datasets, e.g. achieving around 2.4% mAP improvement on COCO, 1.6% on ADE and 1.5% on VG, and the optimal anchors can bring 1.4% to 2.4% mAP improvement on SOTA detectors by only optimizing anchor configurations, e.g. boosting Mask RCNN from 40.3% to 42.3%, and HTC detector from 46.8% to 48.2%.

Linpu Fang, Xingyan Liu, Li Liu et al.

State-of-the-art single depth image-based 3D hand pose estimation methods are based on dense predictions, including voxel-to-voxel predictions, point-to-point regression, and pixel-wise estimations. Despite the good performance, those methods have a few issues in nature, such as the poor trade-off between accuracy and efficiency, and plain feature representation learning with local convolutions. In this paper, a novel pixel-wise prediction-based method is proposed to address the above issues. The key ideas are two-fold: a) explicitly modeling the dependencies among joints and the relations between the pixels and the joints for better local feature representation learning; b) unifying the dense pixel-wise offset predictions and direct joint regression for end-to-end training. Specifically, we first propose a graph convolutional network (GCN) based joint graph reasoning module to model the complex dependencies among joints and augment the representation capability of each pixel. Then we densely estimate all pixels' offsets to joints in both image plane and depth space and calculate the joints' positions by a weighted average over all pixels' predictions, totally discarding the complex postprocessing operations. The proposed model is implemented with an efficient 2D fully convolutional network (FCN) backbone and has only about 1.4M parameters. Extensive experiments on multiple 3D hand pose estimation benchmarks demonstrate that the proposed method achieves new state-of-the-art accuracy while running very efficiently with around a speed of 110fps on a single NVIDIA 1080Ti GPU.

Chenhan Jiang, Shaoju Wang, Hang Xu et al.

Most advances in medical lesion detection network are limited to subtle modification on the conventional detection network designed for natural images. However, there exists a vast domain gap between medical images and natural images where the medical image detection often suffers from several domain-specific challenges, such as high lesion/background similarity, dominant tiny lesions, and severe class imbalance. Is a hand-crafted detection network tailored for natural image undoubtedly good enough over a discrepant medical lesion domain? Is there more powerful operations, filters, and sub-networks that better fit the medical lesion detection problem to be discovered? In this paper, we introduce a novel ElixirNet that includes three components: 1) TruncatedRPN balances positive and negative data for false positive reduction; 2) Auto-lesion Block is automatically customized for medical images to incorporate relation-aware operations among region proposals, and leads to more suitable and efficient classification and localization. 3) Relation transfer module incorporates the semantic relationship and transfers the relevant contextual information with an interpretable the graph thus alleviates the problem of lack of annotations for all types of lesions. Experiments on DeepLesion and Kits19 prove the effectiveness of ElixirNet, achieving improvement of both sensitivity and precision over FPN with fewer parameters.

Linpu Fang, Hang Xu, Zhili Liu et al.

Object detectors trained on fully-annotated data currently yield state of the art performance but require expensive manual annotations. On the other hand, weakly-supervised detectors have much lower performance and cannot be used reliably in a realistic setting. In this paper, we study the hybrid-supervised object detection problem, aiming to train a high quality detector with only a limited amount of fullyannotated data and fully exploiting cheap data with imagelevel labels. State of the art methods typically propose an iterative approach, alternating between generating pseudo-labels and updating a detector. This paradigm requires careful manual hyper-parameter tuning for mining good pseudo labels at each round and is quite time-consuming. To address these issues, we present EHSOD, an end-to-end hybrid-supervised object detection system which can be trained in one shot on both fully and weakly-annotated data. Specifically, based on a two-stage detector, we proposed two modules to fully utilize the information from both kinds of labels: 1) CAMRPN module aims at finding foreground proposals guided by a class activation heat-map; 2) hybrid-supervised cascade module further refines the bounding-box position and classification with the help of an auxiliary head compatible with image-level data. Extensive experiments demonstrate the effectiveness of the proposed method and it achieves comparable results on multiple object detection benchmarks with only 30% fully-annotated data, e.g. 37.5% mAP on COCO. We will release the code and the trained models.

Hang Xu, Linpu Fang, Xiaodan Liang et al.

The dominant object detection approaches treat each dataset separately and fit towards a specific domain, which cannot adapt to other domains without extensive retraining. In this paper, we address the problem of designing a universal object detection model that exploits diverse category granularity from multiple domains and predict all kinds of categories in one system. Existing works treat this problem by integrating multiple detection branches upon one shared backbone network. However, this paradigm overlooks the crucial semantic correlations between multiple domains, such as categories hierarchy, visual similarity, and linguistic relationship. To address these drawbacks, we present a novel universal object detector called Universal-RCNN that incorporates graph transfer learning for propagating relevant semantic information across multiple datasets to reach semantic coherency. Specifically, we first generate a global semantic pool by integrating all high-level semantic representation of all the categories. Then an Intra-Domain Reasoning Module learns and propagates the sparse graph representation within one dataset guided by a spatial-aware GCN. Finally, an InterDomain Transfer Module is proposed to exploit diverse transfer dependencies across all domains and enhance the regional feature representation by attending and transferring semantic contexts globally. Extensive experiments demonstrate that the proposed method significantly outperforms multiple-branch models and achieves the state-of-the-art results on multiple object detection benchmarks (mAP: 49.1% on COCO).

Lewei Yao, Hang Xu, Wei Zhang et al.

The state-of-the-art object detection method is complicated with various modules such as backbone, feature fusion neck, RPN and RCNN head, where each module may have different designs and structures. How to leverage the computational cost and accuracy trade-off for the structural combination as well as the modular selection of multiple modules? Neural architecture search (NAS) has shown great potential in finding an optimal solution. Existing NAS works for object detection only focus on searching better design of a single module such as backbone or feature fusion neck, while neglecting the balance of the whole system. In this paper, we present a two-stage coarse-to-fine searching strategy named Structural-to-Modular NAS (SM-NAS) for searching a GPU-friendly design of both an efficient combination of modules and better modular-level architecture for object detection. Specifically, Structural-level searching stage first aims to find an efficient combination of different modules; Modular-level searching stage then evolves each specific module and pushes the Pareto front forward to a faster task-specific network. We consider a multi-objective search where the search space covers many popular designs of detection methods. We directly search a detection backbone without pre-trained models or any proxy task by exploring a fast training from scratch strategy. The resulting architectures dominate state-of-the-art object detection systems in both inference time and accuracy and demonstrate the effectiveness on multiple detection datasets, e.g. halving the inference time with additional 1% mAP improvement compared to FPN and reaching 46% mAP with the similar inference time of MaskRCNN.

Han Shi, Renjie Pi, Hang Xu et al.

Neural Architecture Search (NAS) has shown great potentials in finding better neural network designs. Sample-based NAS is the most reliable approach which aims at exploring the search space and evaluating the most promising architectures. However, it is computationally very costly. As a remedy, the one-shot approach has emerged as a popular technique for accelerating NAS using weight-sharing. However, due to the weight-sharing of vastly different networks, the one-shot approach is less reliable than the sample-based approach. In this work, we propose BONAS (Bayesian Optimized Neural Architecture Search), a sample-based NAS framework which is accelerated using weight-sharing to evaluate multiple related architectures simultaneously. Specifically, we apply Graph Convolutional Network predictor as a surrogate model for Bayesian Optimization to select multiple related candidate models in each iteration. We then apply weight-sharing to train multiple candidate models simultaneously. This approach not only accelerates the traditional sample-based approach significantly, but also keeps its reliability. This is because weight-sharing among related architectures are more reliable than those in the one-shot approach. Extensive experiments are conducted to verify the effectiveness of our method over many competing algorithms.

Kai Zhang, Shuhang Gu, Radu Timofte et al.

This paper reviews the AIM 2019 challenge on constrained example-based single image super-resolution with focus on proposed solutions and results. The challenge had 3 tracks. Taking the three main aspects (i.e., number of parameters, inference/running time, fidelity (PSNR)) of MSRResNet as the baseline, Track 1 aims to reduce the amount of parameters while being constrained to maintain or improve the running time and the PSNR result, Tracks 2 and 3 aim to optimize running time and PSNR result with constrain of the other two aspects, respectively. Each track had an average of 64 registered participants, and 12 teams submitted the final results. They gauge the state-of-the-art in single image super-resolution.

Vasco Lopes, Fabio Maria Carlucci, Pedro M Esperança et al.

The Neural Architecture Search (NAS) problem is typically formulated as a graph search problem where the goal is to learn the optimal operations over edges in order to maximise a graph-level global objective. Due to the large architecture parameter space, efficiency is a key bottleneck preventing NAS from its practical use. In this paper, we address the issue by framing NAS as a multi-agent problem where agents control a subset of the network and coordinate to reach optimal architectures. We provide two distinct lightweight implementations, with reduced memory requirements (1/8th of state-of-the-art), and performances above those of much more computationally expensive methods. Theoretically, we demonstrate vanishing regrets of the form O(sqrt(T)), with T being the total number of rounds. Finally, aware that random search is an, often ignored, effective baseline we perform additional experiments on 3 alternative datasets and 2 network configurations, and achieve favourable results in comparison.