Feng Liang

Yangguang Li, Bin Huang, Zeren Chen et al.

Recently, perception task based on Bird's-Eye View (BEV) representation has drawn more and more attention, and BEV representation is promising as the foundation for next-generation Autonomous Vehicle (AV) perception. However, most existing BEV solutions either require considerable resources to execute on-vehicle inference or suffer from modest performance. This paper proposes a simple yet effective framework, termed Fast-BEV , which is capable of performing faster BEV perception on the on-vehicle chips. Towards this goal, we first empirically find that the BEV representation can be sufficiently powerful without expensive transformer based transformation nor depth representation. Our Fast-BEV consists of five parts, We novelly propose (1) a lightweight deployment-friendly view transformation which fast transfers 2D image feature to 3D voxel space, (2) an multi-scale image encoder which leverages multi-scale information for better performance, (3) an efficient BEV encoder which is particularly designed to speed up on-vehicle inference. We further introduce (4) a strong data augmentation strategy for both image and BEV space to avoid over-fitting, (5) a multi-frame feature fusion mechanism to leverage the temporal information. Through experiments, on 2080Ti platform, our R50 model can run 52.6 FPS with 47.3% NDS on the nuScenes validation set, exceeding the 41.3 FPS and 47.5% NDS of the BEVDepth-R50 model and 30.2 FPS and 45.7% NDS of the BEVDet4D-R50 model. Our largest model (R101@900x1600) establishes a competitive 53.5% NDS on the nuScenes validation set. We further develop a benchmark with considerable accuracy and efficiency on current popular on-vehicle chips. The code is released at: https://github.com/Sense-GVT/Fast-BEV.

Yufeng Cui, Lichen Zhao, Feng Liang et al.

Contrastive Language-Image Pretraining (CLIP) has emerged as a novel paradigm to learn visual models from language supervision. While researchers continue to push the frontier of CLIP, reproducing these works remains challenging. This is because researchers do not choose consistent training recipes and even use different data, hampering the fair comparison between different methods. In this work, we propose CLIP-benchmark, a first attempt to evaluate, analyze, and benchmark CLIP and its variants. We conduct a comprehensive analysis of three key factors: data, supervision, and model architecture. We find considerable intuitive or counter-intuitive insights: (1). Data quality has a significant impact on performance. (2). Certain supervision has different effects for Convolutional Networks (ConvNets) and Vision Transformers (ViT). Applying more proper supervision can effectively improve the performance of CLIP. (3). Curtailing the text encoder reduces the training cost but not much affect the final performance. Moreover, we further combine DeCLIP with FILIP, bringing us the strongest variant DeFILIP. The CLIP-benchmark would be released at: https://github.com/Sense-GVT/DeCLIP for future CLIP research.

Feng Liang, Bichen Wu, Xiaoliang Dai et al.

Open-vocabulary semantic segmentation aims to segment an image into semantic regions according to text descriptions, which may not have been seen during training. Recent two-stage methods first generate class-agnostic mask proposals and then leverage pre-trained vision-language models, e.g., CLIP, to classify masked regions. We identify the performance bottleneck of this paradigm to be the pre-trained CLIP model, since it does not perform well on masked images. To address this, we propose to finetune CLIP on a collection of masked image regions and their corresponding text descriptions. We collect training data by mining an existing image-caption dataset (e.g., COCO Captions), using CLIP to match masked image regions to nouns in the image captions. Compared with the more precise and manually annotated segmentation labels with fixed classes (e.g., COCO-Stuff), we find our noisy but diverse dataset can better retain CLIP's generalization ability. Along with finetuning the entire model, we utilize the "blank" areas in masked images using a method we dub mask prompt tuning. Experiments demonstrate mask prompt tuning brings significant improvement without modifying any weights of CLIP, and it can further improve a fully finetuned model. In particular, when trained on COCO and evaluated on ADE20K-150, our best model achieves 29.6% mIoU, which is +8.5% higher than the previous state-of-the-art. For the first time, open-vocabulary generalist models match the performance of supervised specialist models in 2017 without dataset-specific adaptations.

Feng Liang, Yangguang Li, Diana Marculescu

Recently, self-supervised Masked Autoencoders (MAE) have attracted unprecedented attention for their impressive representation learning ability. However, the pretext task, Masked Image Modeling (MIM), reconstructs the missing local patches, lacking the global understanding of the image. This paper extends MAE to a fully supervised setting by adding a supervised classification branch, thereby enabling MAE to learn global features from golden labels effectively. The proposed Supervised MAE (SupMAE) only exploits a visible subset of image patches for classification, unlike the standard supervised pre-training where all image patches are used. Through experiments, we demonstrate that SupMAE is not only more training efficient but it also learns more robust and transferable features. Specifically, SupMAE achieves comparable performance with MAE using only 30% of compute when evaluated on ImageNet with the ViT-B/16 model. SupMAE's robustness on ImageNet variants and transfer learning performance outperforms MAE and standard supervised pre-training counterparts. Codes are available at https://github.com/enyac-group/supmae.

Haisheng Fu, Feng Liang, Jie Liang et al.

Recently, deep learning-based image compression has made signifcant progresses, and has achieved better ratedistortion (R-D) performance than the latest traditional method, H.266/VVC, in both subjective metric and the more challenging objective metric. However, a major problem is that many leading learned schemes cannot maintain a good trade-off between performance and complexity. In this paper, we propose an effcient and effective image coding framework, which achieves similar R-D performance with lower complexity than the state of the art. First, we develop an improved multi-scale residual block (MSRB) that can expand the receptive feld and is easier to obtain global information. It can further capture and reduce the spatial correlation of the latent representations. Second, a more advanced importance map network is introduced to adaptively allocate bits to different regions of the image. Third, we apply a 2D post-quantization flter (PQF) to reduce the quantization error, motivated by the Sample Adaptive Offset (SAO) flter in video coding. Moreover, We fnd that the complexity of encoder and decoder have different effects on image compression performance. Based on this observation, we design an asymmetric paradigm, in which the encoder employs three stages of MSRBs to improve the learning capacity, whereas the decoder only needs one stage of MSRB to yield satisfactory reconstruction, thereby reducing the decoding complexity without sacrifcing performance. Experimental results show that compared to the state-of-the-art method, the encoding and decoding time of the proposed method are about 17 times faster, and the R-D performance is only reduced by less than 1% on both Kodak and Tecnick datasets, which is still better than H.266/VVC(4:4:4) and other recent learning-based methods. Our source code is publicly available at https://github.com/fengyurenpingsheng.

Haisheng Fu, Feng Liang

The application of the context-adaptive entropy model significantly improves the rate-distortion (R-D) performance, in which hyperpriors and autoregressive models are jointly utilized to effectively capture the spatial redundancy of the latent representations. However, the latent representations still contain some spatial correlations. In addition, these methods based on the context-adaptive entropy model cannot be accelerated in the decoding process by parallel computing devices, e.g. FPGA or GPU. To alleviate these limitations, we propose a learned multi-resolution image compression framework, which exploits the recently developed octave convolutions to factorize the latent representations into the high-resolution (HR) and low-resolution (LR) parts, similar to wavelet transform, which further improves the R-D performance. To speed up the decoding, our scheme does not use context-adaptive entropy model. Instead, we exploit an additional hyper layer including hyper encoder and hyper decoder to further remove the spatial redundancy of the latent representation. Moreover, the cross-resolution parameter estimation (CRPE) is introduced into the proposed framework to enhance the flow of information and further improve the rate-distortion performance. An additional information-fidelity loss is proposed to the total loss function to adjust the contribution of the LR part to the final bit stream. Experimental results show that our method separately reduces the decoding time by approximately 73.35 % and 93.44 % compared with that of state-of-the-art learned image compression methods, and the R-D performance is still better than H.266/VVC(4:2:0) and some learning-based methods on both PSNR and MS-SSIM metrics across a wide bit rates.

Hung-Yueh Chiang, Natalia Frumkin, Feng Liang et al.

Transfer learning on edge is challenging due to on-device limited resources. Existing work addresses this issue by training a subset of parameters or adding model patches. Developed with inference in mind, Inverted Residual Blocks (IRBs) split a convolutional layer into depthwise and pointwise convolutions, leading to more stacking layers, e.g., convolution, normalization, and activation layers. Though they are efficient for inference, IRBs require that additional activation maps are stored in memory for training weights for convolution layers and scales for normalization layers. As a result, their high memory cost prohibits training IRBs on resource-limited edge devices, and making them unsuitable in the context of transfer learning. To address this issue, we present MobileTL, a memory and computationally efficient on-device transfer learning method for models built with IRBs. MobileTL trains the shifts for internal normalization layers to avoid storing activation maps for the backward pass. Also, MobileTL approximates the backward computation of the activation layer (e.g., Hard-Swish and ReLU6) as a signed function which enables storing a binary mask instead of activation maps for the backward pass. MobileTL fine-tunes a few top blocks (close to output) rather than propagating the gradient through the whole network to reduce the computation cost. Our method reduces memory usage by 46% and 53% for MobileNetV2 and V3 IRBs, respectively. For MobileNetV3, we observe a 36% reduction in floating-point operations (FLOPs) when fine-tuning 5 blocks, while only incurring a 0.6% accuracy reduction on CIFAR10. Extensive experiments on multiple datasets demonstrate that our method is Pareto-optimal (best accuracy under given hardware constraints) compared to prior work in transfer learning for edge devices.

Yang Zhou, Feng Liang, Ting-wu Chin et al.

Machine learning (ML) has entered the mobile era where an enormous number of ML models are deployed on edge devices. However, running common ML models on edge devices continuously may generate excessive heat from the computation, forcing the device to "slow down" to prevent overheating, a phenomenon called thermal throttling. This paper studies the impact of thermal throttling on mobile phones: when it occurs, the CPU clock frequency is reduced, and the model inference latency may increase dramatically. This unpleasant inconsistent behavior has a substantial negative effect on user experience, but it has been overlooked for a long time. To counter thermal throttling, we propose to utilize dynamic networks with shared weights and dynamically shift between large and small ML models seamlessly according to their thermal profile, i.e., shifting to a small model when the system is about to throttle. With the proposed dynamic shifting, the application runs consistently without experiencing CPU clock frequency degradation and latency increase. In addition, we also study the resulting accuracy when dynamic shifting is deployed and show that our approach provides a reasonable trade-off between model latency and model accuracy.

Fenggang Liu, Yangguang Li, Feng Liang et al.

This paper shows that Masking the Deep hierarchical features is an efficient self-supervised method, denoted as MaskDeep. MaskDeep treats each patch in the representation space as an independent instance. We mask part of patches in the representation space and then utilize sparse visible patches to reconstruct high semantic image representation. The intuition of MaskDeep lies in the fact that models can reason from sparse visible patches semantic to the global semantic of the image. We further propose three designs in our framework: 1) a Hierarchical Deep-Masking module to concern the hierarchical property of patch representations, 2) a multi-group strategy to improve the efficiency without any extra computing consumption of the encoder and 3) a multi-target strategy to provide more description of the global semantic. Our MaskDeep brings decent improvements. Trained on ResNet50 with 200 epochs, MaskDeep achieves state-of-the-art results of 71.2% Top1 accuracy linear classification on ImageNet. On COCO object detection tasks, MaskDeep outperforms the self-supervised method SoCo, which specifically designed for object detection. When trained with 100 epochs, MaskDeep achieves 69.6% Top1 accuracy, which surpasses current methods trained with 200 epochs, such as HCSC, by 0.4% .

Shaolun Ruan, Feng Liang, Rohan Ramakrishna et al.

Quantum Neural Networks (QNNs) represent a promising fusion of quantum computing and neural network architectures, offering speed-ups and efficient processing of high-dimensional, entangled data. A crucial component of QNNs is the encoder, which maps classical input data into quantum states. However, choosing suitable encoders remains a significant challenge, largely due to the lack of systematic guidance and the trial-and-error nature of current approaches. This process is further impeded by two key challenges: (1) the difficulty in evaluating encoded quantum states prior to training, and (2) the lack of intuitive methods for analyzing an encoder's ability to effectively distinguish data features. To address these issues, we introduce a novel visualization tool, XQAI-Eyes, which enables QNN developers to compare classical data features with their corresponding encoded quantum states and to examine the mixed quantum states across different classes. By bridging classical and quantum perspectives, XQAI-Eyes facilitates a deeper understanding of how encoders influence QNN performance. Evaluations across diverse datasets and encoder designs demonstrate XQAI-Eyes's potential to support the exploration of the relationship between encoder design and QNN effectiveness, offering a holistic and transparent approach to optimizing quantum encoders. Moreover, domain experts used XQAI-Eyes to derive two key practices for quantum encoder selection, grounded in the principles of pattern preservation and feature mapping.

Feng Liang, Bichen Wu, Jialiang Wang et al.

Diffusion models have transformed the image-to-image (I2I) synthesis and are now permeating into videos. However, the advancement of video-to-video (V2V) synthesis has been hampered by the challenge of maintaining temporal consistency across video frames. This paper proposes a consistent V2V synthesis framework by jointly leveraging spatial conditions and temporal optical flow clues within the source video. Contrary to prior methods that strictly adhere to optical flow, our approach harnesses its benefits while handling the imperfection in flow estimation. We encode the optical flow via warping from the first frame and serve it as a supplementary reference in the diffusion model. This enables our model for video synthesis by editing the first frame with any prevalent I2I models and then propagating edits to successive frames. Our V2V model, FlowVid, demonstrates remarkable properties: (1) Flexibility: FlowVid works seamlessly with existing I2I models, facilitating various modifications, including stylization, object swaps, and local edits. (2) Efficiency: Generation of a 4-second video with 30 FPS and 512x512 resolution takes only 1.5 minutes, which is 3.1x, 7.2x, and 10.5x faster than CoDeF, Rerender, and TokenFlow, respectively. (3) High-quality: In user studies, our FlowVid is preferred 45.7% of the time, outperforming CoDeF (3.5%), Rerender (10.2%), and TokenFlow (40.4%).

Bichen Wu, Ching-Yao Chuang, Xiaoyan Wang et al.

In this paper, we introduce Fairy, a minimalist yet robust adaptation of image-editing diffusion models, enhancing them for video editing applications. Our approach centers on the concept of anchor-based cross-frame attention, a mechanism that implicitly propagates diffusion features across frames, ensuring superior temporal coherence and high-fidelity synthesis. Fairy not only addresses limitations of previous models, including memory and processing speed. It also improves temporal consistency through a unique data augmentation strategy. This strategy renders the model equivariant to affine transformations in both source and target images. Remarkably efficient, Fairy generates 120-frame 512x384 videos (4-second duration at 30 FPS) in just 14 seconds, outpacing prior works by at least 44x. A comprehensive user study, involving 1000 generated samples, confirms that our approach delivers superior quality, decisively outperforming established methods.

Feng Liang, Zhen Zhang, Haifeng Lu et al.

With rapidly increasing distributed deep learning workloads in large-scale data centers, efficient distributed deep learning framework strategies for resource allocation and workload scheduling have become the key to high-performance deep learning. The large-scale environment with large volumes of datasets, models, and computational and communication resources raises various unique challenges for resource allocation and workload scheduling in distributed deep learning, such as scheduling complexity, resource and workload heterogeneity, and fault tolerance. To uncover these challenges and corresponding solutions, this survey reviews the literature, mainly from 2019 to 2024, on efficient resource allocation and workload scheduling strategies for large-scale distributed DL. We explore these strategies by focusing on various resource types, scheduling granularity levels, and performance goals during distributed training and inference processes. We highlight critical challenges for each topic and discuss key insights of existing technologies. To illustrate practical large-scale resource allocation and workload scheduling in real distributed deep learning scenarios, we use a case study of training large language models. This survey aims to encourage computer science, artificial intelligence, and communications researchers to understand recent advances and explore future research directions for efficient framework strategies for large-scale distributed deep learning.

Luya Wang, Feng Liang, Yangguang Li et al.

Recently, self-supervised vision transformers have attracted unprecedented attention for their impressive representation learning ability. However, the dominant method, contrastive learning, mainly relies on an instance discrimination pretext task, which learns a global understanding of the image. This paper incorporates local feature learning into self-supervised vision transformers via Reconstructive Pre-training (RePre). Our RePre extends contrastive frameworks by adding a branch for reconstructing raw image pixels in parallel with the existing contrastive objective. RePre is equipped with a lightweight convolution-based decoder that fuses the multi-hierarchy features from the transformer encoder. The multi-hierarchy features provide rich supervisions from low to high semantic information, which are crucial for our RePre. Our RePre brings decent improvements on various contrastive frameworks with different vision transformer architectures. Transfer performance in downstream tasks outperforms supervised pre-training and state-of-the-art (SOTA) self-supervised counterparts.

Yinyin Chen, Shishuang He, Yun Yang et al.

Topic models provide a useful text-mining tool for learning, extracting, and discovering latent structures in large text corpora. Although a plethora of methods have been proposed for topic modeling, lacking in the literature is a formal theoretical investigation of the statistical identifiability and accuracy of latent topic estimation. In this paper, we propose a maximum likelihood estimator (MLE) of latent topics based on a specific integrated likelihood that is naturally connected to the concept, in computational geometry, of volume minimization. Our theory introduces a new set of geometric conditions for topic model identifiability, conditions that are weaker than conventional separability conditions, which typically rely on the existence of pure topic documents or of anchor words. Weaker conditions allow a wider and thus potentially more fruitful investigation. We conduct finite-sample error analysis for the proposed estimator and discuss connections between our results and those of previous investigations. We conclude with empirical studies employing both simulated and real datasets.

Daniel Khashabi, Mojtaba Ziyadi, Feng Liang

In this work we propose a heteroscedastic generalization to RVM, a fast Bayesian framework for regression, based on some recent similar works. We use variational approximation and expectation propagation to tackle the problem. The work is still under progress and we are examining the results and comparing with the previous works.