This paper proposes a hardware-efficient architecture, Linearized Convolution Network (LiCo-Net) for keyword spotting. It is optimized specifically for low-power processor units like microcontrollers. ML operators exhibit heterogeneous efficiency profiles on power-efficient hardware. Given the exact theoretical computation cost, int8 operators are more computation-effective than float operators, and linear layers are often more efficient than other layers. The proposed LiCo-Net is a dual-phase system that uses the efficient int8 linear operators at the inference phase and applies streaming convolutions at the training phase to maintain a high model capacity. The experimental results show that LiCo-Net outperforms single-value decomposition filter (SVDF) on hardware efficiency with on-par detection performance. Compared to SVDF, LiCo-Net reduces cycles by 40% on HiFi4 DSP.
HDR(High Dynamic Range) video can reproduce realistic scenes more realistically, with a wider gamut and broader brightness range. HDR video resources are still scarce, and most videos are still stored in SDR (Standard Dynamic Range) format. Therefore, SDRTV-to-HDRTV Conversion (SDR video to HDR video) can significantly enhance the user's video viewing experience. Since the correlation between adjacent video frames is very high, the method utilizing the information of multiple frames can improve the quality of the converted HDRTV. Therefore, we propose a multi-frame fusion neural network \textbf{DSLNet} for SDRTV to HDRTV conversion. We first propose a dynamic spatial-temporal feature alignment module \textbf{DMFA}, which can align and fuse multi-frame. Then a novel spatial-temporal feature modulation module \textbf{STFM}, STFM extracts spatial-temporal information of adjacent frames for more accurate feature modulation. Finally, we design a quality enhancement module \textbf{LKQE} with large kernels, which can enhance the quality of generated HDR videos. To evaluate the performance of the proposed method, we construct a corresponding multi-frame dataset using HDR video of the HDR10 standard to conduct a comprehensive evaluation of different methods. The experimental results show that our method obtains state-of-the-art performance. The dataset and code will be released.
Joint super-resolution and inverse tone-mapping (SR-ITM) aims to enhance the visual quality of videos that have quality deficiencies in resolution and dynamic range. This problem arises when using 4K high dynamic range (HDR) TVs to watch a low-resolution standard dynamic range (LR SDR) video. Previous methods that rely on learning local information typically cannot do well in preserving color conformity and long-range structural similarity, resulting in unnatural color transition and texture artifacts. In order to tackle these challenges, we propose a global priors guided modulation network (GPGMNet) for joint SR-ITM. In particular, we design a global priors extraction module (GPEM) to extract color conformity prior and structural similarity prior that are beneficial for ITM and SR tasks, respectively. To further exploit the global priors and preserve spatial information, we devise multiple global priors guided spatial-wise modulation blocks (GSMBs) with a few parameters for intermediate feature modulation, in which the modulation parameters are generated by the shared global priors and the spatial features map from the spatial pyramid convolution block (SPCB). With these elaborate designs, the GPGMNet can achieve higher visual quality with lower computational complexity. Extensive experiments demonstrate that our proposed GPGMNet is superior to the state-of-the-art methods. Specifically, our proposed model exceeds the state-of-the-art by 0.64 dB in PSNR, with 69$\%$ fewer parameters and 3.1$\times$ speedup. The code will be released soon.
Acoustic events are sounds with well-defined spectro-temporal characteristics which can be associated with the physical objects generating them. Acoustic scenes are collections of such acoustic events in no specific temporal order. Given this natural linkage between events and scenes, a common belief is that the ability to classify events must help in the classification of scenes. This has led to several efforts attempting to do well on Acoustic Event Tagging (AET) and Acoustic Scene Classification (ASC) using a multi-task network. However, in these efforts, improvement in one task does not guarantee an improvement in the other, suggesting a tension between ASC and AET. It is unclear if improvements in AET translates to improvements in ASC. We explore this conundrum through an extensive empirical study and show that under certain conditions, using AET as an auxiliary task in the multi-task network consistently improves ASC performance. Additionally, ASC performance further improves with the AET data-set size and is not sensitive to the choice of events or the number of events in the AET data-set. We conclude that this improvement in ASC performance comes from the regularization effect of using AET and not from the network's improved ability to discern between acoustic events.
This paper reviews the challenge on constrained high dynamic range (HDR) imaging that was part of the New Trends in Image Restoration and Enhancement (NTIRE) workshop, held in conjunction with CVPR 2022. This manuscript focuses on the competition set-up, datasets, the proposed methods and their results. The challenge aims at estimating an HDR image from multiple respective low dynamic range (LDR) observations, which might suffer from under- or over-exposed regions and different sources of noise. The challenge is composed of two tracks with an emphasis on fidelity and complexity constraints: In Track 1, participants are asked to optimize objective fidelity scores while imposing a low-complexity constraint (i.e. solutions can not exceed a given number of operations). In Track 2, participants are asked to minimize the complexity of their solutions while imposing a constraint on fidelity scores (i.e. solutions are required to obtain a higher fidelity score than the prescribed baseline). Both tracks use the same data and metrics: Fidelity is measured by means of PSNR with respect to a ground-truth HDR image (computed both directly and with a canonical tonemapping operation), while complexity metrics include the number of Multiply-Accumulate (MAC) operations and runtime (in seconds).
Standard acoustic event classification (AEC) solutions require large-scale collection of data from client devices for model optimization. Federated learning (FL) is a compelling framework that decouples data collection and model training to enhance customer privacy. In this work, we investigate the feasibility of applying FL to improve AEC performance while no customer data can be directly uploaded to the server. We assume no pseudo labels can be inferred from on-device user inputs, aligning with the typical use cases of AEC. We adapt self-supervised learning to the FL framework for on-device continual learning of representations, and it results in improved performance of the downstream AEC classifiers without labeled/pseudo-labeled data available. Compared to the baseline w/o FL, the proposed method improves precision up to 20.3\% relatively while maintaining the recall. Our work differs from prior work in FL in that our approach does not require user-generated learning targets, and the data we use is collected from our Beta program and is de-identified, to maximally simulate the production settings.
Intelligent service robots require the ability to perform a variety of tasks in dynamic environments. Despite the significant progress in robotic grasping, it is still a challenge for robots to decide grasping position when given different tasks in unstructured real life environments. In order to overcome this challenge, creating a proper knowledge representation framework is the key. Unlike the previous work, in this paper, task is defined as a triplet including grasping tool, desired action and target object. Our proposed algorithm GATER (Grasp--Action--Target Embeddings and Relations) models the relationship among grasping tools--action--target objects in embedding space. To validate our method, a novel dataset is created for task-specific grasping. GATER is trained on the new dataset and achieve task-specific grasping inference with 94.6\% success rate. Finally, the effectiveness of GATER algorithm is tested on a real service robot platform. GATER algorithm has its potential in human behavior prediction and human-robot interaction.
We present BN-NAS, neural architecture search with Batch Normalization (BN-NAS), to accelerate neural architecture search (NAS). BN-NAS can significantly reduce the time required by model training and evaluation in NAS. Specifically, for fast evaluation, we propose a BN-based indicator for predicting subnet performance at a very early training stage. The BN-based indicator further facilitates us to improve the training efficiency by only training the BN parameters during the supernet training. This is based on our observation that training the whole supernet is not necessary while training only BN parameters accelerates network convergence for network architecture search. Extensive experiments show that our method can significantly shorten the time of training supernet by more than 10 times and shorten the time of evaluating subnets by more than 600,000 times without losing accuracy.
In this paper, we observe two levels of redundancies when applying vision transformers (ViT) for image recognition. First, fixing the number of tokens through the whole network produces redundant features at the spatial level. Second, the attention maps among different transformer layers are redundant. Based on the observations above, we propose a PSViT: a ViT with token Pooling and attention Sharing to reduce the redundancy, effectively enhancing the feature representation ability, and achieving a better speed-accuracy trade-off. Specifically, in our PSViT, token pooling can be defined as the operation that decreases the number of tokens at the spatial level. Besides, attention sharing will be built between the neighboring transformer layers for reusing the attention maps having a strong correlation among adjacent layers. Then, a compact set of the possible combinations for different token pooling and attention sharing mechanisms are constructed. Based on the proposed compact set, the number of tokens in each layer and the choices of layers sharing attention can be treated as hyper-parameters that are learned from data automatically. Experimental results show that the proposed scheme can achieve up to 6.6% accuracy improvement in ImageNet classification compared with the DeiT.
We introduce the first Neural Architecture Search (NAS) method to find a better transformer architecture for image recognition. Recently, transformers without CNN-based backbones are found to achieve impressive performance for image recognition. However, the transformer is designed for NLP tasks and thus could be sub-optimal when directly used for image recognition. In order to improve the visual representation ability for transformers, we propose a new search space and searching algorithm. Specifically, we introduce a locality module that models the local correlations in images explicitly with fewer computational cost. With the locality module, our search space is defined to let the search algorithm freely trade off between global and local information as well as optimizing the low-level design choice in each module. To tackle the problem caused by huge search space, a hierarchical neural architecture search method is proposed to search the optimal vision transformer from two levels separately with the evolutionary algorithm. Extensive experiments on the ImageNet dataset demonstrate that our method can find more discriminative and efficient transformer variants than the ResNet family (e.g., ResNet101) and the baseline ViT for image classification.