Tao Chen

Wenqiang Zhang, Zilong Huang, Guozhong Luo et al. · deepmind, tencent-ai

Although vision transformers (ViTs) have achieved great success in computer vision, the heavy computational cost hampers their applications to dense prediction tasks such as semantic segmentation on mobile devices. In this paper, we present a mobile-friendly architecture named \textbf{To}ken \textbf{P}yramid Vision Trans\textbf{former} (\textbf{TopFormer}). The proposed \textbf{TopFormer} takes Tokens from various scales as input to produce scale-aware semantic features, which are then injected into the corresponding tokens to augment the representation. Experimental results demonstrate that our method significantly outperforms CNN- and ViT-based networks across several semantic segmentation datasets and achieves a good trade-off between accuracy and latency. On the ADE20K dataset, TopFormer achieves 5\% higher accuracy in mIoU than MobileNetV3 with lower latency on an ARM-based mobile device. Furthermore, the tiny version of TopFormer achieves real-time inference on an ARM-based mobile device with competitive results. The code and models are available at: https://github.com/hustvl/TopFormer

Jiakang Yuan, Bo Zhang, Xiangchao Yan et al. · deepmind

Unsupervised Domain Adaptation (UDA) technique has been explored in 3D cross-domain tasks recently. Though preliminary progress has been made, the performance gap between the UDA-based 3D model and the supervised one trained with fully annotated target domain is still large. This motivates us to consider selecting partial-yet-important target data and labeling them at a minimum cost, to achieve a good trade-off between high performance and low annotation cost. To this end, we propose a Bi-domain active learning approach, namely Bi3D, to solve the cross-domain 3D object detection task. The Bi3D first develops a domainness-aware source sampling strategy, which identifies target-domain-like samples from the source domain to avoid the model being interfered by irrelevant source data. Then a diversity-based target sampling strategy is developed, which selects the most informative subset of target domain to improve the model adaptability to the target domain using as little annotation budget as possible. Experiments are conducted on typical cross-domain adaptation scenarios including cross-LiDAR-beam, cross-country, and cross-sensor, where Bi3D achieves a promising target-domain detection accuracy (89.63% on KITTI) compared with UDAbased work (84.29%), even surpassing the detector trained on the full set of the labeled target domain (88.98%). Our code is available at: https://github.com/PJLabADG/3DTrans.

Chuangguan Ye, Hongyuan Zhu, Bo Zhang et al. · deepmind

In recent years, research on few-shot learning (FSL) has been fast-growing in the 2D image domain due to the less requirement for labeled training data and greater generalization for novel classes. However, its application in 3D point cloud data is relatively under-explored. Not only need to distinguish unseen classes as in the 2D domain, 3D FSL is more challenging in terms of irregular structures, subtle inter-class differences, and high intra-class variances {when trained on a low number of data.} Moreover, different architectures and learning algorithms make it difficult to study the effectiveness of existing 2D FSL algorithms when migrating to the 3D domain. In this work, for the first time, we perform systematic and extensive investigations of directly applying recent 2D FSL works to 3D point cloud related backbone networks and thus suggest a strong learning baseline for few-shot 3D point cloud classification. Furthermore, we propose a new network, Point-cloud Correlation Interaction (PCIA), with three novel plug-and-play components called Salient-Part Fusion (SPF) module, Self-Channel Interaction Plus (SCI+) module, and Cross-Instance Fusion Plus (CIF+) module to obtain more representative embeddings and improve the feature distinction. These modules can be inserted into most FSL algorithms with minor changes and significantly improve the performance. Experimental results on three benchmark datasets, ModelNet40-FS, ShapeNet70-FS, and ScanObjectNN-FS, demonstrate that our method achieves state-of-the-art performance for the 3D FSL task. Code and datasets are available at https://github.com/cgye96/A_Closer_Look_At_3DFSL.

Sijin Chen, Hongyuan Zhu, Xin Chen et al. · deepmind

3D dense captioning aims to generate multiple captions localized with their associated object regions. Existing methods follow a sophisticated ``detect-then-describe'' pipeline equipped with numerous hand-crafted components. However, these hand-crafted components would yield suboptimal performance given cluttered object spatial and class distributions among different scenes. In this paper, we propose a simple-yet-effective transformer framework Vote2Cap-DETR based on recent popular \textbf{DE}tection \textbf{TR}ansformer (DETR). Compared with prior arts, our framework has several appealing advantages: 1) Without resorting to numerous hand-crafted components, our method is based on a full transformer encoder-decoder architecture with a learnable vote query driven object decoder, and a caption decoder that produces the dense captions in a set-prediction manner. 2) In contrast to the two-stage scheme, our method can perform detection and captioning in one-stage. 3) Without bells and whistles, extensive experiments on two commonly used datasets, ScanRefer and Nr3D, demonstrate that our Vote2Cap-DETR surpasses current state-of-the-arts by 11.13\% and 7.11\% in CIDEr@0.5IoU, respectively. Codes will be released soon.

Yongqi Huang, Peng Ye, Xiaoshui Huang et al. · deepmind

Structural re-parameterization is a general training scheme for Convolutional Neural Networks (CNNs), which achieves performance improvement without increasing inference cost. As Vision Transformers (ViTs) are gradually surpassing CNNs in various visual tasks, one may question: if a training scheme specifically for ViTs exists that can also achieve performance improvement without increasing inference cost? Recently, Mixture-of-Experts (MoE) has attracted increasing attention, as it can efficiently scale up the capacity of Transformers at a fixed cost through sparsely activated experts. Considering that MoE can also be viewed as a multi-branch structure, can we utilize MoE to implement a ViT training scheme similar to structural re-parameterization? In this paper, we affirmatively answer these questions, with a new general training strategy for ViTs. Specifically, we decouple the training and inference phases of ViTs. During training, we replace some Feed-Forward Networks (FFNs) of the ViT with specially designed, more efficient MoEs that assign tokens to experts by random uniform partition, and perform Experts Weights Averaging (EWA) on these MoEs at the end of each iteration. After training, we convert each MoE into an FFN by averaging the experts, transforming the model back into original ViT for inference. We further provide a theoretical analysis to show why and how it works. Comprehensive experiments across various 2D and 3D visual tasks, ViT architectures, and datasets validate the effectiveness and generalizability of the proposed training scheme. Besides, our training scheme can also be applied to improve performance when fine-tuning ViTs. Lastly, but equally important, the proposed EWA technique can significantly improve the effectiveness of naive MoE in various 2D visual small datasets and 3D visual tasks.

Peng Ye, Tong He, Baopu Li et al. · deepmind

Neural Architecture Search has attracted increasing attention in recent years. Among them, differential NAS approaches such as DARTS, have gained popularity for the search efficiency. However, they still suffer from three main issues, that are, the weak stability due to the performance collapse, the poor generalization ability of the searched architectures, and the inferior robustness to different kinds of proxies. To solve the stability and generalization problems, a simple-but-effective regularization method, termed as Beta-Decay, is proposed to regularize the DARTS-based NAS searching process (i.e., $β$-DARTS). Specifically, Beta-Decay regularization can impose constraints to keep the value and variance of activated architecture parameters from being too large, thereby ensuring fair competition among architecture parameters and making the supernet less sensitive to the impact of input on the operation set. In-depth theoretical analyses on how it works and why it works are provided. Comprehensive experiments validate that Beta-Decay regularization can help to stabilize the searching process and makes the searched network more transferable across different datasets. To address the robustness problem, we first benchmark different NAS methods under a wide range of proxy data, proxy channels, proxy layers and proxy epochs, since the robustness of NAS under different kinds of proxies has not been explored before. We then conclude some interesting findings and find that $β$-DARTS always achieves the best result among all compared NAS methods under almost all proxies. We further introduce the novel flooding regularization to the weight optimization of $β$-DARTS (i.e., Bi-level regularization), and experimentally and theoretically verify its effectiveness for improving the proxy robustness of differentiable NAS.

Chongshan Lu, Fukun Yin, Xin Chen et al. · deepmind

Neural Radiance Fields (NeRF) has achieved impressive results in single object scene reconstruction and novel view synthesis, which have been demonstrated on many single modality and single object focused indoor scene datasets like DTU, BMVS, and NeRF Synthetic.However, the study of NeRF on large-scale outdoor scene reconstruction is still limited, as there is no unified outdoor scene dataset for large-scale NeRF evaluation due to expensive data acquisition and calibration costs. In this paper, we propose a large-scale outdoor multi-modal dataset, OMMO dataset, containing complex land objects and scenes with calibrated images, point clouds and prompt annotations. Meanwhile, a new benchmark for several outdoor NeRF-based tasks is established, such as novel view synthesis, surface reconstruction, and multi-modal NeRF. To create the dataset, we capture and collect a large number of real fly-view videos and select high-quality and high-resolution clips from them. Then we design a quality review module to refine images, remove low-quality frames and fail-to-calibrate scenes through a learning-based automatic evaluation plus manual review. Finally, a number of volunteers are employed to add the text descriptions for each scene and key-frame to meet the potential multi-modal requirements in the future. Compared with existing NeRF datasets, our dataset contains abundant real-world urban and natural scenes with various scales, camera trajectories, and lighting conditions. Experiments show that our dataset can benchmark most state-of-the-art NeRF methods on different tasks. We will release the dataset and model weights very soon.

Shengji Tang, Peng Ye, Baopu Li et al.

Recent research understands the residual networks from a new perspective of the implicit ensemble model. From this view, previous methods such as stochastic depth and stimulative training have further improved the performance of the residual network by sampling and training of its subnets. However, they both use the same supervision for all subnets of different capacities and neglect the valuable knowledge generated by subnets during training. In this manuscript, we mitigate the significant knowledge distillation gap caused by using the same kind of supervision and advocate leveraging the subnets to provide diverse knowledge. Based on this motivation, we propose a group knowledge based training framework for boosting the performance of residual networks. Specifically, we implicitly divide all subnets into hierarchical groups by subnet-in-subnet sampling, aggregate the knowledge of different subnets in each group during training, and exploit upper-level group knowledge to supervise lower-level subnet groups. Meanwhile, We also develop a subnet sampling strategy that naturally samples larger subnets, which are found to be more helpful than smaller subnets in boosting performance for hierarchical groups. Compared with typical subnet training and other methods, our method achieves the best efficiency and performance trade-offs on multiple datasets and network structures. The code is at https://github.com/tsj-001/AAAI24-GKT.

Bo Zhang, Xinyu Cai, Jiakang Yuan et al. · deepmind

Domain shifts such as sensor type changes and geographical situation variations are prevalent in Autonomous Driving (AD), which poses a challenge since AD model relying on the previous domain knowledge can be hardly directly deployed to a new domain without additional costs. In this paper, we provide a new perspective and approach of alleviating the domain shifts, by proposing a Reconstruction-Simulation-Perception (ReSimAD) scheme. Specifically, the implicit reconstruction process is based on the knowledge from the previous old domain, aiming to convert the domain-related knowledge into domain-invariant representations, e.g., 3D scene-level meshes. Besides, the point clouds simulation process of multiple new domains is conditioned on the above reconstructed 3D meshes, where the target-domain-like simulation samples can be obtained, thus reducing the cost of collecting and annotating new-domain data for the subsequent perception process. For experiments, we consider different cross-domain situations such as Waymo-to-KITTI, Waymo-to-nuScenes, Waymo-to-ONCE, etc, to verify the zero-shot target-domain perception using ReSimAD. Results demonstrate that our method is beneficial to boost the domain generalization ability, even promising for 3D pre-training.

Bo Zhang, Tao Chen, Bin Wang

Although domain adaptation has been extensively studied in natural image-based segmentation task, the research on cross-domain segmentation for very high resolution (VHR) remote sensing images (RSIs) still remains underexplored. The VHR RSIs-based cross-domain segmentation mainly faces two critical challenges: 1) Large area land covers with many diverse object categories bring severe local patch-level data distribution deviations, thus yielding different adaptation difficulties for different local patches; 2) Different VHR sensor types or dynamically changing modes cause the VHR images to go through intensive data distribution differences even for the same geographical location, resulting in different global feature-level domain gap. To address these challenges, we propose a curriculum-style local-to-global cross-domain adaptation framework for the segmentation of VHR RSIs. The proposed curriculum-style adaptation performs the adaptation process in an easy-to-hard way according to the adaptation difficulties that can be obtained using an entropy-based score for each patch of the target domain, and thus well aligns the local patches in a domain image. The proposed local-to-global adaptation performs the feature alignment process from the locally semantic to globally structural feature discrepancies, and consists of a semantic-level domain classifier and an entropy-level domain classifier that can reduce the above cross-domain feature discrepancies. Extensive experiments have been conducted in various cross-domain scenarios, including geographic location variations and imaging mode variations, and the experimental results demonstrate that the proposed method can significantly boost the domain adaptability of segmentation networks for VHR RSIs. Our code is available at: https://github.com/BOBrown/CCDA_LGFA.

Yancheng Cai, Bo Zhang, Baopu Li et al. · cambridge, deepmind

Cross-domain pedestrian detection aims to generalize pedestrian detectors from one label-rich domain to another label-scarce domain, which is crucial for various real-world applications. Most recent works focus on domain alignment to train domain-adaptive detectors either at the instance level or image level. From a practical point of view, one-stage detectors are faster. Therefore, we concentrate on designing a cross-domain algorithm for rapid one-stage detectors that lacks instance-level proposals and can only perform image-level feature alignment. However, pure image-level feature alignment causes the foreground-background misalignment issue to arise, i.e., the foreground features in the source domain image are falsely aligned with background features in the target domain image. To address this issue, we systematically analyze the importance of foreground and background in image-level cross-domain alignment, and learn that background plays a more critical role in image-level cross-domain alignment. Therefore, we focus on cross-domain background feature alignment while minimizing the influence of foreground features on the cross-domain alignment stage. This paper proposes a novel framework, namely, background-focused distribution alignment (BFDA), to train domain adaptive onestage pedestrian detectors. Specifically, BFDA first decouples the background features from the whole image feature maps and then aligns them via a novel long-short-range discriminator.

Weimin Wu, Jiayuan Fan, Tao Chen et al. · deepmind

The linear ensemble based strategy, i.e., averaging ensemble, has been proposed to improve the performance in unsupervised domain adaptation tasks. However, a typical UDA task is usually challenged by dynamically changing factors, such as variable weather, views, and background in the unlabeled target domain. Most previous ensemble strategies ignore UDA's dynamic and uncontrollable challenge, facing limited feature representations and performance bottlenecks. To enhance the model, adaptability between domains and reduce the computational cost when deploying the ensemble model, we propose a novel framework, namely Instance aware Model Ensemble With Distillation, IMED, which fuses multiple UDA component models adaptively according to different instances and distills these components into a small model. The core idea of IMED is a dynamic instance aware ensemble strategy, where for each instance, a nonlinear fusion subnetwork is learned that fuses the extracted features and predicted labels of multiple component models. The nonlinear fusion method can help the ensemble model handle dynamically changing factors. After learning a large capacity ensemble model with good adaptability to different changing factors, we leverage the ensemble teacher model to guide the learning of a compact student model by knowledge distillation. Furthermore, we provide the theoretical analysis of the validity of IMED for UDA. Extensive experiments conducted on various UDA benchmark datasets, e.g., Office 31, Office Home, and VisDA 2017, show the superiority of the model based on IMED to the state of the art methods under the comparable computation cost.

Hancheng Ye, Bo Zhang, Tao Chen et al.

Global channel pruning (GCP) aims to remove a subset of channels (filters) across different layers from a deep model without hurting the performance. Previous works focus on either single task model pruning or simply adapting it to multitask scenario, and still face the following problems when handling multitask pruning: 1) Due to the task mismatch, a well-pruned backbone for classification task focuses on preserving filters that can extract category-sensitive information, causing filters that may be useful for other tasks to be pruned during the backbone pruning stage; 2) For multitask predictions, different filters within or between layers are more closely related and interacted than that for single task prediction, making multitask pruning more difficult. Therefore, aiming at multitask model compression, we propose a Performance-Aware Global Channel Pruning (PAGCP) framework. We first theoretically present the objective for achieving superior GCP, by considering the joint saliency of filters from intra- and inter-layers. Then a sequentially greedy pruning strategy is proposed to optimize the objective, where a performance-aware oracle criterion is developed to evaluate sensitivity of filters to each task and preserve the globally most task-related filters. Experiments on several multitask datasets show that the proposed PAGCP can reduce the FLOPs and parameters by over 60% with minor performance drop, and achieves 1.2x$\sim$3.3x acceleration on both cloud and mobile platforms.

Decheng Liu, Tao Chen, Chunlei Peng et al.

Due to the successful development of deep image generation technology, visual data forgery detection would play a more important role in social and economic security. Existing forgery detection methods suffer from unsatisfactory generalization ability to determine the authenticity in the unseen domain. In this paper, we propose a novel Attention Consistency Refined masked frequency forgery representation model toward generalizing face forgery detection algorithm (ACMF). Most forgery technologies always bring in high-frequency aware cues, which make it easy to distinguish source authenticity but difficult to generalize to unseen artifact types. The masked frequency forgery representation module is designed to explore robust forgery cues by randomly discarding high-frequency information. In addition, we find that the forgery attention map inconsistency through the detection network could affect the generalizability. Thus, the forgery attention consistency is introduced to force detectors to focus on similar attention regions for better generalization ability. Experiment results on several public face forgery datasets (FaceForensic++, DFD, Celeb-DF, and WDF datasets) demonstrate the superior performance of the proposed method compared with the state-of-the-art methods.

Zhen Wang, Xinyun Jiang, Jun Xiao et al.

Explicit Caption Editing (ECE) -- refining reference image captions through a sequence of explicit edit operations (e.g., KEEP, DETELE) -- has raised significant attention due to its explainable and human-like nature. After training with carefully designed reference and ground-truth caption pairs, state-of-the-art ECE models exhibit limited generalization ability beyond the original training data distribution, i.e., they are tailored to refine content details only in in-domain samples but fail to correct errors in out-of-domain samples. To this end, we propose a new Diffusion-based Explicit Caption editing method: DECap. Specifically, we reformulate the ECE task as a denoising process under the diffusion mechanism, and introduce innovative edit-based noising and denoising processes. Thanks to this design, the noising process can help to eliminate the need for meticulous paired data selection by directly introducing word-level noises for training, learning diverse distribution over input reference caption. The denoising process involves the explicit predictions of edit operations and corresponding content words, refining reference captions through iterative step-wise editing. To further efficiently implement our diffusion process and improve the inference speed, DECap discards the prevalent multi-stage design and directly generates edit operations and content words simultaneously. Extensive ablations have demonstrated the strong generalization ability of DECap in various scenarios. More interestingly, it even shows great potential in improving the quality and controllability of caption generation.

Yiying Yang, Wen Liu, Fukun Yin et al.

Recent advancements in implicit neural representations have contributed to high-fidelity surface reconstruction and photorealistic novel view synthesis. However, the computational complexity inherent in these methodologies presents a substantial impediment, constraining the attainable frame rates and resolutions in practical applications. In response to this predicament, we propose VQ-NeRF, an effective and efficient pipeline for enhancing implicit neural representations via vector quantization. The essence of our method involves reducing the sampling space of NeRF to a lower resolution and subsequently reinstating it to the original size utilizing a pre-trained VAE decoder, thereby effectively mitigating the sampling time bottleneck encountered during rendering. Although the codebook furnishes representative features, reconstructing fine texture details of the scene remains challenging due to high compression rates. To overcome this constraint, we design an innovative multi-scale NeRF sampling scheme that concurrently optimizes the NeRF model at both compressed and original scales to enhance the network's ability to preserve fine details. Furthermore, we incorporate a semantic loss function to improve the geometric fidelity and semantic coherence of our 3D reconstructions. Extensive experiments demonstrate the effectiveness of our model in achieving the optimal trade-off between rendering quality and efficiency. Evaluation on the DTU, BlendMVS, and H3DS datasets confirms the superior performance of our approach.

Weihao Lin, Tao Chen, Chong Yu

Semi-supervised video object segmentation (Semi-VOS), which requires only annotating the first frame of a video to segment future frames, has received increased attention recently. Among existing pipelines, the memory-matching-based one is becoming the main research stream, as it can fully utilize the temporal sequence information to obtain high-quality segmentation results. Even though this type of method has achieved promising performance, the overall framework still suffers from heavy computation overhead, mainly caused by the per-frame dense convolution operations between high-resolution feature maps and each kernel filter. Therefore, we propose a sparse baseline of VOS named SpVOS in this work, which develops a novel triple sparse convolution to reduce the computation costs of the overall VOS framework. The designed triple gate, taking full consideration of both spatial and temporal redundancy between adjacent video frames, adaptively makes a triple decision to decide how to apply the sparse convolution on each pixel to control the computation overhead of each layer, while maintaining sufficient discrimination capability to distinguish similar objects and avoid error accumulation. A mixed sparse training strategy, coupled with a designed objective considering the sparsity constraint, is also developed to balance the VOS segmentation performance and computation costs. Experiments are conducted on two mainstream VOS datasets, including DAVIS and Youtube-VOS. Results show that, the proposed SpVOS achieves superior performance over other state-of-the-art sparse methods, and even maintains comparable performance, e.g., an 83.04% (79.29%) overall score on the DAVIS-2017 (Youtube-VOS) validation set, with the typical non-sparse VOS baseline (82.88% for DAVIS-2017 and 80.36% for Youtube-VOS) while saving up to 42% FLOPs, showing its application potential for resource-constrained scenarios.

Mingsheng Li, Lin Zhang, Mingzhen Zhu et al.

This paper studies the problem of pre-training for small models, which is essential for many mobile devices. Current state-of-the-art methods on this problem transfer the representational knowledge of a large network (as a Teacher) into a smaller model (as a Student) using self-supervised distillation, improving the performance of the small model on downstream tasks. However, existing approaches are insufficient in extracting the crucial knowledge that is useful for discerning categories in downstream tasks during the distillation process. In this paper, for the first time, we introduce language guidance to the distillation process and propose a new method named Language-Guided Distillation (LGD) system, which uses category names of the target downstream task to help refine the knowledge transferred between the teacher and student. To this end, we utilize a pre-trained text encoder to extract semantic embeddings from language and construct a textual semantic space called Textual Semantics Bank (TSB). Furthermore, we design a Language-Guided Knowledge Aggregation (LGKA) module to construct the visual semantic space, also named Visual Semantics Bank (VSB). The task-related knowledge is transferred by driving a student encoder to mimic the similarity score distribution inferred by a teacher over TSB and VSB. Compared with other small models obtained by either ImageNet pre-training or self-supervised distillation, experiment results show that the distilled lightweight model using the proposed LGD method presents state-of-the-art performance and is validated on various downstream tasks, including classification, detection, and segmentation. We have made the code available at https://github.com/mZhenz/LGD.

Jingjing Chen, Jichao Zhang, Enver Sangineto et al.

Gaze redirection aims at manipulating the gaze of a given face image with respect to a desired direction (i.e., a reference angle) and it can be applied to many real life scenarios, such as video-conferencing or taking group photos. However, previous work on this topic mainly suffers of two limitations: (1) Low-quality image generation and (2) Low redirection precision. In this paper, we propose to alleviate these problems by means of a novel gaze redirection framework which exploits both a numerical and a pictorial direction guidance, jointly with a coarse-to-fine learning strategy. Specifically, the coarse branch learns the spatial transformation which warps input image according to desired gaze. On the other hand, the fine-grained branch consists of a generator network with conditional residual image learning and a multi-task discriminator. This second branch reduces the gap between the previously warped image and the ground-truth image and recovers finer texture details. Moreover, we propose a numerical and pictorial guidance module~(NPG) which uses a pictorial gazemap description and numerical angles as an extra guide to further improve the precision of gaze redirection. Extensive experiments on a benchmark dataset show that the proposed method outperforms the state-of-the-art approaches in terms of both image quality and redirection precision. The code is available at https://github.com/jingjingchen777/CFGR

Shengze Li, Jianjian Cao, Peng Ye et al.

Recently, foundational models such as CLIP and SAM have shown promising performance for the task of Zero-Shot Anomaly Segmentation (ZSAS). However, either CLIP-based or SAM-based ZSAS methods still suffer from non-negligible key drawbacks: 1) CLIP primarily focuses on global feature alignment across different inputs, leading to imprecise segmentation of local anomalous parts; 2) SAM tends to generate numerous redundant masks without proper prompt constraints, resulting in complex post-processing requirements. In this work, we innovatively propose a CLIP and SAM collaboration framework called ClipSAM for ZSAS. The insight behind ClipSAM is to employ CLIP's semantic understanding capability for anomaly localization and rough segmentation, which is further used as the prompt constraints for SAM to refine the anomaly segmentation results. In details, we introduce a crucial Unified Multi-scale Cross-modal Interaction (UMCI) module for interacting language with visual features at multiple scales of CLIP to reason anomaly positions. Then, we design a novel Multi-level Mask Refinement (MMR) module, which utilizes the positional information as multi-level prompts for SAM to acquire hierarchical levels of masks and merges them. Extensive experiments validate the effectiveness of our approach, achieving the optimal segmentation performance on the MVTec-AD and VisA datasets.

Peng Ye, Yongqi Huang, Chongjun Tu et al.

Fine-tuning pre-trained foundation models has gained significant popularity in various research fields. Existing methods for fine-tuning can be roughly divided into two categories, namely Parameter-Efficient Fine-Tuning and High-Performance Fine-Tuning. The former aims at improving efficiency, while the latter focuses on enhancing performance. Beyond these methods, we demonstrate that Partial Fine-Tuning can be an innovative and promising direction capable of concurrently enhancing both efficiency and accuracy. We first validate eight manually-defined partial fine-tuning strategies across kinds of datasets and vision transformer architectures, and find that some partial fine-tuning strategies (e.g., ffn only or attention only) can achieve better performance with fewer tuned parameters than full fine-tuning, and selecting appropriate layers is critical to partial fine-tuning. Thus, we propose a novel fine-tuned angle metric to guide the selection of appropriate layers for partial fine-tuning, making it flexible to be adapted to various scenarios for more practicable partial fine-tuning. Additionally, we show that partial fine-tuning can serve as a new dimension for Model Soups, improving both the model performance and generalization with fewer tuned parameters. Comprehensive experiments on a wide range of datasets and models validate the great potential of partial fine-tuning.

Tianshuo Peng, Mingsheng Li, Jiakang Yuan et al.

Recent advancements in Large Multi-modal Models (LMMs) underscore the importance of scaling by increasing image-text paired data, achieving impressive performance on general tasks. Despite their effectiveness in broad applications, generalist models are primarily trained on web-scale datasets dominated by natural images, resulting in the sacrifice of specialized capabilities for domain-specific tasks that require extensive domain prior knowledge. Moreover, directly integrating expert models tailored for specific domains is challenging due to the representational gap and imbalanced optimization between the generalist model and experts. To address these challenges, we introduce Chimera, a scalable and low-cost multi-modal pipeline designed to boost the ability of existing LMMs with domain-specific experts. Specifically, we design a progressive training strategy to integrate features from expert models into the input of a generalist LMM. To address the imbalanced optimization caused by the well-aligned general visual encoder, we introduce a novel Generalist-Specialist Collaboration Masking (GSCM) mechanism. This results in a versatile model that excels across the chart, table, math, and document domains, achieving state-of-the-art performance on multi-modal reasoning and visual content extraction tasks, both of which are challenging tasks for assessing existing LMMs.

Maosen Zhao, Pengtao Chen, Chong Yu et al.

Model quantization reduces the bit-width of weights and activations, improving memory efficiency and inference speed in diffusion models. However, achieving 4-bit quantization remains challenging. Existing methods, primarily based on integer quantization and post-training quantization fine-tuning, struggle with inconsistent performance. Inspired by the success of floating-point (FP) quantization in large language models, we explore low-bit FP quantization for diffusion models and identify key challenges: the failure of signed FP quantization to handle asymmetric activation distributions, the insufficient consideration of temporal complexity in the denoising process during fine-tuning, and the misalignment between fine-tuning loss and quantization error. To address these challenges, we propose the mixup-sign floating-point quantization (MSFP) framework, first introducing unsigned FP quantization in model quantization, along with timestep-aware LoRA (TALoRA) and denoising-factor loss alignment (DFA), which ensure precise and stable fine-tuning. Extensive experiments show that we are the first to achieve superior performance in 4-bit FP quantization for diffusion models, outperforming existing PTQ fine-tuning methods in 4-bit INT quantization.

Chong Yu, Tao Chen, Zhongxue Gan et al.

The transformer extends its success from the language to the vision domain. Because of the stacked self-attention and cross-attention blocks, the acceleration deployment of vision transformer on GPU hardware is challenging and also rarely studied. This paper thoroughly designs a compression scheme to maximally utilize the GPU-friendly 2:4 fine-grained structured sparsity and quantization. Specially, an original large model with dense weight parameters is first pruned into a sparse one by 2:4 structured pruning, which considers the GPU's acceleration of 2:4 structured sparse pattern with FP16 data type, then the floating-point sparse model is further quantized into a fixed-point one by sparse-distillation-aware quantization aware training, which considers GPU can provide an extra speedup of 2:4 sparse calculation with integer tensors. A mixed-strategy knowledge distillation is used during the pruning and quantization process. The proposed compression scheme is flexible to support supervised and unsupervised learning styles. Experiment results show GPUSQ-ViT scheme achieves state-of-the-art compression by reducing vision transformer models 6.4-12.7 times on model size and 30.3-62 times on FLOPs with negligible accuracy degradation on ImageNet classification, COCO detection and ADE20K segmentation benchmarking tasks. Moreover, GPUSQ-ViT can boost actual deployment performance by 1.39-1.79 times and 3.22-3.43 times of latency and throughput on A100 GPU, and 1.57-1.69 times and 2.11-2.51 times improvement of latency and throughput on AGX Orin.

Bicheng Guo, Tao Chen, Shibo He et al.

Neural Architecture Search (NAS) is a powerful tool for automating effective image processing DNN designing. The ranking has been advocated to design an efficient performance predictor for NAS. The previous contrastive method solves the ranking problem by comparing pairs of architectures and predicting their relative performance. However, it only focuses on the rankings between two involved architectures and neglects the overall quality distributions of the search space, which may suffer generalization issues. A predictor, namely Neural Architecture Ranker (NAR) which concentrates on the global quality tier of specific architecture, is proposed to tackle such problems caused by the local perspective. The NAR explores the quality tiers of the search space globally and classifies each individual to the tier they belong to according to its global ranking. Thus, the predictor gains the knowledge of the performance distributions of the search space which helps to generalize its ranking ability to the datasets more easily. Meanwhile, the global quality distribution facilitates the search phase by directly sampling candidates according to the statistics of quality tiers, which is free of training a search algorithm, e.g., Reinforcement Learning (RL) or Evolutionary Algorithm (EA), thus it simplifies the NAS pipeline and saves the computational overheads. The proposed NAR achieves better performance than the state-of-the-art methods on two widely used datasets for NAS research. On the vast search space of NAS-Bench-101, the NAR easily finds the architecture with top 0.01$\unicode{x2030}$ performance only by sampling. It also generalizes well to different image datasets of NAS-Bench-201, i.e., CIFAR-10, CIFAR-100, and ImageNet-16-120 by identifying the optimal architectures for each of them.

Haichao Zhang, Gang Yu, Tao Chen et al.

Video creation has been an attractive yet challenging task for artists to explore. With the advancement of deep learning, recent works try to utilize deep convolutional neural networks to synthesize a video with the aid of a guiding video, and have achieved promising results. However, the acquisition of guiding videos, or other forms of guiding temporal information is costly expensive and difficult in reality. Therefore, in this work we introduce a new video synthesis task by employing two rough bad-drwan sketches only as input to create a realistic portrait video. A two-stage Sketch-to-Video model is proposed, which consists of two key novelties: 1) a feature retrieve and projection (FRP) module, which parititions the input sketch into different parts and utilizes these parts for synthesizing a realistic start or end frame and meanwhile generating rich semantic features, is designed to alleviate the sketch out-of-domain problem due to arbitrarily drawn free-form sketch styles by different users. 2) A motion projection followed by feature blending module, which projects a video (used only in training phase) into a motion space modeled by normal distribution and blends the motion variables with semantic features extracted above, is proposed to alleviate the guiding temporal information missing problem in the test phase. Experiments conducted on a combination of CelebAMask-HQ and VoxCeleb2 dataset well validate that, our method can acheive both good quantitative and qualitative results in synthesizing high-quality videos from two rough bad-drawn sketches.

Haichao Zhang, Youcheng Ben, Weixi Zhang et al.

Recent face reenactment works are limited by the coarse reference landmarks, leading to unsatisfactory identity preserving performance due to the distribution gap between the manipulated landmarks and those sampled from a real person. To address this issue, we propose a fine-grained identity-preserving landmark-guided face reenactment approach. The proposed method has two novelties. First, a landmark synthesis network which is designed to generate fine-grained landmark faces with more details. The network refines the manipulated landmarks and generates a smooth and gradually changing face landmark sequence with good identity preserving ability. Second, several novel loss functions including synthesized face identity preserving loss, foreground/background mask loss as well as boundary loss are designed, which aims at synthesizing clear and sharp high-quality faces. Experiments are conducted on our self-collected BeautySelfie and the public VoxCeleb1 datasets. The presented qualitative and quantitative results show that our method can reenact fine-grained higher quality faces with good ID-preserved appearance details, fewer artifacts and clearer boundaries than state-of-the-art works. Code will be released for reproduction.

Yike Wu, Bo Zhang, Gang Yu et al.

The goal of few-shot fine-grained image classification is to recognize rarely seen fine-grained objects in the query set, given only a few samples of this class in the support set. Previous works focus on learning discriminative image features from a limited number of training samples for distinguishing various fine-grained classes, but ignore one important fact that spatial alignment of the discriminative semantic features between the query image with arbitrary changes and the support image, is also critical for computing the semantic similarity between each support-query pair. In this work, we propose an object-aware long-short-range spatial alignment approach, which is composed of a foreground object feature enhancement (FOE) module, a long-range semantic correspondence (LSC) module and a short-range spatial manipulation (SSM) module. The FOE is developed to weaken background disturbance and encourage higher foreground object response. To address the problem of long-range object feature misalignment between support-query image pairs, the LSC is proposed to learn the transferable long-range semantic correspondence by a designed feature similarity metric. Further, the SSM module is developed to refine the transformed support feature after the long-range step to align short-range misaligned features (or local details) with the query features. Extensive experiments have been conducted on four benchmark datasets, and the results show superior performance over most state-of-the-art methods under both 1-shot and 5-shot classification scenarios.

Xuelin Qian, Huazhu Fu, Weiya Shi et al.

To counter the outbreak of COVID-19, the accurate diagnosis of suspected cases plays a crucial role in timely quarantine, medical treatment, and preventing the spread of the pandemic. Considering the limited training cases and resources (e.g, time and budget), we propose a Multi-task Multi-slice Deep Learning System (M3Lung-Sys) for multi-class lung pneumonia screening from CT imaging, which only consists of two 2D CNN networks, i.e., slice- and patient-level classification networks. The former aims to seek the feature representations from abundant CT slices instead of limited CT volumes, and for the overall pneumonia screening, the latter one could recover the temporal information by feature refinement and aggregation between different slices. In addition to distinguish COVID-19 from Healthy, H1N1, and CAP cases, our M 3 Lung-Sys also be able to locate the areas of relevant lesions, without any pixel-level annotation. To further demonstrate the effectiveness of our model, we conduct extensive experiments on a chest CT imaging dataset with a total of 734 patients (251 healthy people, 245 COVID-19 patients, 105 H1N1 patients, and 133 CAP patients). The quantitative results with plenty of metrics indicate the superiority of our proposed model on both slice- and patient-level classification tasks. More importantly, the generated lesion location maps make our system interpretable and more valuable to clinicians.

Rongliang Wu, Gongjie Zhang, Shijian Lu et al.

Recent advances in Generative Adversarial Nets (GANs) have shown remarkable improvements for facial expression editing. However, current methods are still prone to generate artifacts and blurs around expression-intensive regions, and often introduce undesired overlapping artifacts while handling large-gap expression transformations such as transformation from furious to laughing. To address these limitations, we propose Cascade Expression Focal GAN (Cascade EF-GAN), a novel network that performs progressive facial expression editing with local expression focuses. The introduction of the local focus enables the Cascade EF-GAN to better preserve identity-related features and details around eyes, noses and mouths, which further helps reduce artifacts and blurs within the generated facial images. In addition, an innovative cascade transformation strategy is designed by dividing a large facial expression transformation into multiple small ones in cascade, which helps suppress overlapping artifacts and produce more realistic editing while dealing with large-gap expression transformations. Extensive experiments over two publicly available facial expression datasets show that our proposed Cascade EF-GAN achieves superior performance for facial expression editing.

Wenxuan Wang, Yanwei Fu, Qiang Sun et al.

Affective computing and cognitive theory are widely used in modern human-computer interaction scenarios. Human faces, as the most prominent and easily accessible features, have attracted great attention from researchers. Since humans have rich emotions and developed musculature, there exist a lot of fine-grained expressions in real-world applications. However, it is extremely time-consuming to collect and annotate a large number of facial images, of which may even require psychologists to correctly categorize them. To the best of our knowledge, the existing expression datasets are only limited to several basic facial expressions, which are not sufficient to support our ambitions in developing successful human-computer interaction systems. To this end, a novel Fine-grained Facial Expression Database - F2ED is contributed in this paper, and it includes more than 200k images with 54 facial expressions from 119 persons. Considering the phenomenon of uneven data distribution and lack of samples is common in real-world scenarios, we further evaluate several tasks of few-shot expression learning by virtue of our F2ED, which are to recognize the facial expressions given only few training instances. These tasks mimic human performance to learn robust and general representation from few examples. To address such few-shot tasks, we propose a unified task-driven framework - Compositional Generative Adversarial Network (Comp-GAN) learning to synthesize facial images and thus augmenting the instances of few-shot expression classes. Extensive experiments are conducted on F2ED and existing facial expression datasets, i.e., JAFFE and FER2013, to validate the efficacy of our F2ED in pre-training facial expression recognition network and the effectiveness of our proposed approach Comp-GAN to improve the performance of few-shot recognition tasks.

Wenxuan Wang, Qiang Sun, Tao Chen et al.

The recent research of facial expression recognition has made a lot of progress due to the development of deep learning technologies, but some typical challenging problems such as the variety of rich facial expressions and poses are still not resolved. To solve these problems, we develop a new Facial Expression Recognition (FER) framework by involving the facial poses into our image synthesizing and classification process. There are two major novelties in this work. First, we create a new facial expression dataset of more than 200k images with 119 persons, 4 poses and 54 expressions. To our knowledge this is the first dataset to label faces with subtle emotion changes for expression recognition purpose. It is also the first dataset that is large enough to validate the FER task on unbalanced poses, expressions, and zero-shot subject IDs. Second, we propose a facial pose generative adversarial network (FaPE-GAN) to synthesize new facial expression images to augment the data set for training purpose, and then learn a LightCNN based Fa-Net model for expression classification. Finally, we advocate four novel learning tasks on this dataset. The experimental results well validate the effectiveness of the proposed approach.

Feng Zhou, Shu Kong, Charless Fowlkes et al.

Automated facial expression analysis has a variety of applications in human-computer interaction. Traditional methods mainly analyze prototypical facial expressions of no more than eight discrete emotions as a classification task. However, in practice, spontaneous facial expressions in naturalistic environment can represent not only a wide range of emotions, but also different intensities within an emotion family. In such situation, these methods are not reliable or adequate. In this paper, we propose to train deep convolutional neural networks (CNNs) to analyze facial expressions explainable in a dimensional emotion model. The proposed method accommodates not only a set of basic emotion expressions, but also a full range of other emotions and subtle emotion intensities that we both feel in ourselves and perceive in others in our daily life. Specifically, we first mapped facial expressions into dimensional measures so that we transformed facial expression analysis from a classification problem to a regression one. We then tested our CNN-based methods for facial expression regression and these methods demonstrated promising performance. Moreover, we improved our method by a bilinear pooling which encodes second-order statistics of features. We showed such bilinear-CNN models significantly outperformed their respective baselines.