Prosodic boundary plays an important role in text-to-speech synthesis (TTS) in terms of naturalness and readability. However, the acquisition of prosodic boundary labels relies on manual annotation, which is costly and time-consuming. In this paper, we propose to automatically extract prosodic boundary labels from text-audio data via a neural text-speech model with pre-trained audio encoders. This model is pre-trained on text and speech data separately and jointly fine-tuned on TTS data in a triplet format: {speech, text, prosody}. The experimental results on both automatic evaluation and human evaluation demonstrate that: 1) the proposed text-speech prosody annotation framework significantly outperforms text-only baselines; 2) the quality of automatic prosodic boundary annotations is comparable to human annotations; 3) TTS systems trained with model-annotated boundaries are slightly better than systems that use manual ones.
In the past years, learned image compression (LIC) has achieved remarkable performance. The recent LIC methods outperform VVC in both PSNR and MS-SSIM. However, the low bit-rate reconstructions of LIC suffer from artifacts such as blurring, color drifting and texture missing. Moreover, those varied artifacts make image quality metrics correlate badly with human perceptual quality. In this paper, we propose PO-ELIC, i.e., Perception-Oriented Efficient Learned Image Coding. To be specific, we adapt ELIC, one of the state-of-the-art LIC models, with adversarial training techniques. We apply a mixture of losses including hinge-form adversarial loss, Charbonnier loss, and style loss, to finetune the model towards better perceptual quality. Experimental results demonstrate that our method achieves comparable perceptual quality with HiFiC with much lower bitrate.
In recent years, sequential recommender systems (SRSs) and session-based recommender systems (SBRSs) have emerged as a new paradigm of RSs to capture users' short-term but dynamic preferences for enabling more timely and accurate recommendations. Although SRSs and SBRSs have been extensively studied, there are many inconsistencies in this area caused by the diverse descriptions, settings, assumptions and application domains. There is no work to provide a unified framework and problem statement to remove the commonly existing and various inconsistencies in the area of SR/SBR. There is a lack of work to provide a comprehensive and systematic demonstration of the data characteristics, key challenges, most representative and state-of-the-art approaches, typical real-world applications and important future research directions in the area. This work aims to fill in these gaps so as to facilitate further research in this exciting and vibrant area.
With the growth of transport modes, high traffic forecasting precision is required in intelligent transportation systems. Most previous works utilize the transformer architecture based on graph neural networks and attention mechanisms to discover spatiotemporal dependencies and dynamic relationships. The correlation information among spatiotemporal sequences, however, has not been thoroughly considered. In this paper, we present two elaborate spatiotemporal representations, spatial correlation information (SCorr) and temporal correlation information (TCorr), among spatiotemporal sequences based on the maximal information coefficient. Using SCorr, we propose a novel correlation information-based spatiotemporal network (CorrSTN), including a dynamic graph neural network component incorporating correlation information into the spatial structure effectively and a multi-head attention component utilizing spatial correlation information to extract dynamic temporal dependencies accurately. Using TCorr, we further explore the correlation pattern among different periodic data and then propose a novel data selection scheme to identify the most relevant data. The experimental results on the highway traffic flow (PEMS07 and PEMS08) and metro crowd flow (HZME inflow and outflow) datasets demonstrate that CorrSTN outperforms the state-of-the-art methods in terms of predictive performance. In particular, on the HZME (outflow) dataset, our model makes significant improvements compared with the latest model ASTGNN by 12.7%, 14.4% and 27.4% in the metrics of MAE, RMSE and MAPE, respectively.
This paper reviews the NTIRE 2022 challenge on efficient single image super-resolution with focus on the proposed solutions and results. The task of the challenge was to super-resolve an input image with a magnification factor of $\times$4 based on pairs of low and corresponding high resolution images. The aim was to design a network for single image super-resolution that achieved improvement of efficiency measured according to several metrics including runtime, parameters, FLOPs, activations, and memory consumption while at least maintaining the PSNR of 29.00dB on DIV2K validation set. IMDN is set as the baseline for efficiency measurement. The challenge had 3 tracks including the main track (runtime), sub-track one (model complexity), and sub-track two (overall performance). In the main track, the practical runtime performance of the submissions was evaluated. The rank of the teams were determined directly by the absolute value of the average runtime on the validation set and test set. In sub-track one, the number of parameters and FLOPs were considered. And the individual rankings of the two metrics were summed up to determine a final ranking in this track. In sub-track two, all of the five metrics mentioned in the description of the challenge including runtime, parameter count, FLOPs, activations, and memory consumption were considered. Similar to sub-track one, the rankings of five metrics were summed up to determine a final ranking. The challenge had 303 registered participants, and 43 teams made valid submissions. They gauge the state-of-the-art in efficient single image super-resolution.
It is not accurate to make recommendations only based one single current session. Therefore, multi-session-based recommendation(MSBR) is a solution for the problem. Compared with the previous MSBR models, we have made three improvements in this paper. First, the previous work choose to use all the history sessions of the user and/or of his similar users. When the user's current interest changes greatly from the past, most of these sessions can only have negative impacts. Therefore, we select a large number of randomly chosen sessions from the dataset as candidate sessions to avoid over depending on history data. Then we only choose to use the most similar sessions to get the most useful information while reduce the noise caused by dissimilar sessions. Second, in real-world datasets, short sessions account for a large proportion. The RNN often used in previous work is not suitable to process short sessions, because RNN only focuses on the sequential relationship, which we find is not the only relationship between items in short sessions. So, we designed a more suitable method named GAFE based on attention to process short sessions. Third, Although there are few long sessions, they can not be ignored. Not like previous models, which simply process long sessions in the same way as short sessions, we propose LSIS, which can split the interest of long sessions, to make better use of long sessions. Finally, to help recommendations, we also have considered users' long-term interests captured by a multi-layer GRU. Considering the four points above, we built the model ENIREC. Experiments on two real-world datasets show that the comprehensive performance of ENIREC is better than other existing models.
Generative language models (LMs) such as GPT-2/3 can be prompted to generate text with remarkable quality. While they are designed for text-prompted generation, it remains an open question how the generation process could be guided by modalities beyond text such as images. In this work, we propose a training-free framework, called MAGIC (iMAge-Guided text generatIon with CLIP), for plugging in visual controls in the generation process and enabling LMs to perform multimodal tasks (e.g., image captioning) in a zero-shot manner. MAGIC is a simple yet efficient plug-and-play framework, which directly combines an off-the-shelf LM (i.e., GPT-2) and an image-text matching model (i.e., CLIP) for image-grounded text generation. During decoding, MAGIC influences the generation of the LM by introducing a CLIP-induced score, called magic score, which regularizes the generated result to be semantically related to a given image while being coherent to the previously generated context. Notably, the proposed decoding scheme does not involve any gradient update operation, therefore being computationally efficient. On the challenging task of zero-shot image captioning, MAGIC outperforms the state-of-the-art method by notable margins with a nearly 27 times decoding speedup. MAGIC is a flexible framework and is theoretically compatible with any text generation tasks that incorporate image grounding. In the experiments, we showcase that it is also capable of performing visually grounded story generation given both an image and a text prompt.
With the recently massive development in convolution neural networks, numerous lightweight CNN-based image super-resolution methods have been proposed for practical deployments on edge devices. However, most existing methods focus on one specific aspect: network or loss design, which leads to the difficulty of minimizing the model size. To address the issue, we conclude block devising, architecture searching, and loss design to obtain a more efficient SR structure. In this paper, we proposed an edge-enhanced feature distillation network, named EFDN, to preserve the high-frequency information under constrained resources. In detail, we build an edge-enhanced convolution block based on the existing reparameterization methods. Meanwhile, we propose edge-enhanced gradient loss to calibrate the reparameterized path training. Experimental results show that our edge-enhanced strategies preserve the edge and significantly improve the final restoration quality. Code is available at https://github.com/icandle/EFDN.
Despite the exciting performance, Transformer is criticized for its excessive parameters and computation cost. However, compressing Transformer remains as an open problem due to its internal complexity of the layer designs, i.e., Multi-Head Attention (MHA) and Feed-Forward Network (FFN). To address this issue, we introduce Group-wise Transformation towards a universal yet lightweight Transformer for vision-and-language tasks, termed as LW-Transformer. LW-Transformer applies Group-wise Transformation to reduce both the parameters and computations of Transformer, while also preserving its two main properties, i.e., the efficient attention modeling on diverse subspaces of MHA, and the expanding-scaling feature transformation of FFN. We apply LW-Transformer to a set of Transformer-based networks, and quantitatively measure them on three vision-and-language tasks and six benchmark datasets. Experimental results show that while saving a large number of parameters and computations, LW-Transformer achieves very competitive performance against the original Transformer networks for vision-and-language tasks. To examine the generalization ability, we also apply our optimization strategy to a recently proposed image Transformer called Swin-Transformer for image classification, where the effectiveness can be also confirmed
JPEG is a popular image compression method widely used by individuals, data center, cloud storage and network filesystems. However, most recent progress on image compression mainly focuses on uncompressed images while ignoring trillions of already-existing JPEG images. To compress these JPEG images adequately and restore them back to JPEG format losslessly when needed, we propose a deep learning based JPEG recompression method that operates on DCT domain and propose a Multi-Level Cross-Channel Entropy Model to compress the most informative Y component. Experiments show that our method achieves state-of-the-art performance compared with traditional JPEG recompression methods including Lepton, JPEG XL and CMIX. To the best of our knowledge, this is the first learned compression method that losslessly transcodes JPEG images to more storage-saving bitstreams.