Large-scale Causal Language Models (CLMs), e.g., GPT3 and ChatGPT, have brought great success in text generation. However, it is still an open challenge to control the generation process of CLM while balancing flexibility, control granularity, and generation efficiency. In this paper, we provide a new alternative for controllable text generation (CTG), by designing a non-intrusive, lightweight control plugin to accompany the generation of CLM at arbitrary time steps. The proposed control plugin, namely Residual Memory Transformer (RMT), has an encoder-decoder setup, which can accept any types of control conditions and cooperate with CLM through a residual learning paradigm, to achieve a more flexible, general, and efficient CTG. Extensive experiments are carried out on various control tasks, in the form of both automatic and human evaluations. The results show the superiority of RMT over a range of state-of-the-art approaches, proving the effectiveness and versatility of our approach.
Understanding semantic content from brain activity during sleep represents a major goal in neuroscience. While studies in rodents have shown spontaneous neural reactivation of memories during sleep, capturing the semantic content of human sleep poses a significant challenge due to the absence of well-annotated sleep datasets and the substantial differences in neural patterns between wakefulness and sleep. To address these challenges, we designed a novel cognitive neuroscience experiment and collected a comprehensive, well-annotated electroencephalography (EEG) dataset from 134 subjects during both wakefulness and sleep. Leveraging this benchmark dataset, we developed SI-SD that enhances sleep semantic decoding through the position-wise alignment of neural latent sequence between wakefulness and sleep. In the 15-way classification task, our model achieves 24.12% and 21.39% top-1 accuracy on unseen subjects for NREM 2/3 and REM sleep, respectively, surpassing all other baselines. With additional fine-tuning, decoding performance improves to 30.32% and 31.65%, respectively. Besides, inspired by previous neuroscientific findings, we systematically analyze how the "Slow Oscillation" event impacts decoding performance in NREM 2/3 sleep -- decoding performance on unseen subjects further improves to 40.02%. Together, our findings and methodologies contribute to a promising neuro-AI framework for decoding brain activity during sleep.
We introduce Transformer-VQ, a decoder-only transformer computing softmax-based dense self-attention in linear time. Transformer-VQ's efficient attention is enabled by vector-quantized keys and a novel caching mechanism. In our large-scale experiments, Transformer-VQ is shown highly competitive in quality, obtaining 0.99 bpb on Enwik8, 26.6 ppl on PG-19, and 3.16 bpb on ImageNet64. In addition, the optimized implementation of Transformer-VQ is over 3x faster than a comparable quadratic-time transformer at sequence length 8k, is over 12x faster at 32k, and can scale to 131k with similar throughput. Code available: \url{https://github.com/transformer-vq/transformer_vq}
Query-based object detectors have made significant advancements since the publication of DETR. However, most existing methods still rely on multi-stage encoders and decoders, or a combination of both. Despite achieving high accuracy, the multi-stage paradigm (typically consisting of 6 stages) suffers from issues such as heavy computational burden, prompting us to reconsider its necessity. In this paper, we explore multiple techniques to enhance query-based detectors and, based on these findings, propose a novel model called GOLO (Global Once and Local Once), which follows a two-stage decoding paradigm. Compared to other mainstream query-based models with multi-stage decoders, our model employs fewer decoder stages while still achieving considerable performance. Experimental results on the COCO dataset demonstrate the effectiveness of our approach.
In this technical report, we present a solution for 3D object generation of ICCV 2023 OmniObject3D Challenge. In recent years, 3D object generation has made great process and achieved promising results, but it remains a challenging task due to the difficulty of generating complex, textured and high-fidelity results. To resolve this problem, we study learning effective NeRFs and SDFs representations with 3D Generative Adversarial Networks (GANs) for 3D object generation. Specifically, inspired by recent works, we use the efficient geometry-aware 3D GANs as the backbone incorporating with label embedding and color mapping, which enables to train the model on different taxonomies simultaneously. Then, through a decoder, we aggregate the resulting features to generate Neural Radiance Fields (NeRFs) based representations for rendering high-fidelity synthetic images. Meanwhile, we optimize Signed Distance Functions (SDFs) to effectively represent objects with 3D meshes. Besides, we observe that this model can be effectively trained with only a few images of each object from a variety of classes, instead of using a great number of images per object or training one model per class. With this pipeline, we can optimize an effective model for 3D object generation. This solution is one of the final top-3-place solutions in the ICCV 2023 OmniObject3D Challenge.
Recent works in end-to-end speech-to-text translation (ST) have proposed multi-tasking methods with soft parameter sharing which leverage machine translation (MT) data via secondary encoders that map text inputs to an eventual cross-modal representation. In this work, we instead propose a ST/MT multi-tasking framework with hard parameter sharing in which all model parameters are shared cross-modally. Our method reduces the speech-text modality gap via a pre-processing stage which converts speech and text inputs into two discrete token sequences of similar length -- this allows models to indiscriminately process both modalities simply using a joint vocabulary. With experiments on MuST-C, we demonstrate that our multi-tasking framework improves attentional encoder-decoder, Connectionist Temporal Classification (CTC), transducer, and joint CTC/attention models by an average of +0.5 BLEU without any external MT data. Further, we show that this framework incorporates external MT data, yielding +0.8 BLEU, and also improves transfer learning from pre-trained textual models, yielding +1.8 BLEU.
Voice conversion is becoming increasingly popular, and a growing number of application scenarios require models with streaming inference capabilities. The recently proposed DualVC attempts to achieve this objective through streaming model architecture design and intra-model knowledge distillation along with hybrid predictive coding to compensate for the lack of future information. However, DualVC encounters several problems that limit its performance. First, the autoregressive decoder has error accumulation in its nature and limits the inference speed as well. Second, the causal convolution enables streaming capability but cannot sufficiently use future information within chunks. Third, the model is unable to effectively address the noise in the unvoiced segments, lowering the sound quality. In this paper, we propose DualVC 2 to address these issues. Specifically, the model backbone is migrated to a Conformer-based architecture, empowering parallel inference. Causal convolution is replaced by non-causal convolution with dynamic chunk mask to make better use of within-chunk future information. Also, quiet attention is introduced to enhance the model's noise robustness. Experiments show that DualVC 2 outperforms DualVC and other baseline systems in both subjective and objective metrics, with only 186.4 ms latency. Our audio samples are made publicly available.
This paper addresses the problem of estimating the 3-DoF camera pose for a ground-level image with respect to a satellite image that encompasses the local surroundings. We propose a novel end-to-end approach that leverages the learning of dense pixel-wise flow fields in pairs of ground and satellite images to calculate the camera pose. Our approach differs from existing methods by constructing the feature metric at the pixel level, enabling full-image supervision for learning distinctive geometric configurations and visual appearances across views. Specifically, our method employs two distinct convolution networks for ground and satellite feature extraction. Then, we project the ground feature map to the bird's eye view (BEV) using a fixed camera height assumption to achieve preliminary geometric alignment. To further establish content association between the BEV and satellite features, we introduce a residual convolution block to refine the projected BEV feature. Optical flow estimation is performed on the refined BEV feature map and the satellite feature map using flow decoder networks based on RAFT. After obtaining dense flow correspondences, we apply the least square method to filter matching inliers and regress the ground camera pose. Extensive experiments demonstrate significant improvements compared to state-of-the-art methods. Notably, our approach reduces the median localization error by 89%, 19%, 80% and 35% on the KITTI, Ford multi-AV, VIGOR and Oxford RobotCar datasets, respectively.
Traffic forecasting, crucial for urban planning, requires accurate predictions of spatial-temporal traffic patterns across urban areas. Existing research mainly focuses on designing complex models that capture spatial-temporal dependencies among variables explicitly. However, this field faces challenges related to data scarcity and model stability, which results in limited performance improvement. To address these issues, we propose Spatial-Temporal Masked AutoEncoders (STMAE), a plug-and-play framework designed to enhance existing spatial-temporal models on traffic prediction. STMAE consists of two learning stages. In the pretraining stage, an encoder processes partially visible traffic data produced by a dual-masking strategy, including biased random walk-based spatial masking and patch-based temporal masking. Subsequently, two decoders aim to reconstruct the masked counterparts from both spatial and temporal perspectives. The fine-tuning stage retains the pretrained encoder and integrates it with decoders from existing backbones to improve forecasting accuracy. Our results on traffic benchmarks show that STMAE can largely enhance the forecasting capabilities of various spatial-temporal models.
Attention-based encoder-decoder models with autoregressive (AR) decoding have proven to be the dominant approach for automatic speech recognition (ASR) due to their superior accuracy. However, they often suffer from slow inference. This is primarily attributed to the incremental calculation of the decoder. This work proposes a partially AR framework, which employs segment-level vectorized beam search for improving the inference speed of an ASR model based on the hybrid connectionist temporal classification (CTC) attention-based architecture. It first generates an initial hypothesis using greedy CTC decoding, identifying low-confidence tokens based on their output probabilities. We then utilize the decoder to perform segment-level vectorized beam search on these tokens, re-predicting in parallel with minimal decoder calculations. Experimental results show that our method is 12 to 13 times faster in inference on the LibriSpeech corpus over AR decoding whilst preserving high accuracy.
Semi-supervised learning (SSL) has been proven beneficial for mitigating the issue of limited labeled data especially on the task of volumetric medical image segmentation. Unlike previous SSL methods which focus on exploring highly confident pseudo-labels or developing consistency regularization schemes, our empirical findings suggest that inconsistent decoder features emerge naturally when two decoders strive to generate consistent predictions. Based on the observation, we first analyze the treasure of discrepancy in learning towards consistency, under both pseudo-labeling and consistency regularization settings, and subsequently propose a novel SSL method called LeFeD, which learns the feature-level discrepancy obtained from two decoders, by feeding the discrepancy as a feedback signal to the encoder. The core design of LeFeD is to enlarge the difference by training differentiated decoders, and then learn from the inconsistent information iteratively. We evaluate LeFeD against eight state-of-the-art (SOTA) methods on three public datasets. Experiments show LeFeD surpasses competitors without any bells and whistles such as uncertainty estimation and strong constraints, as well as setting a new state-of-the-art for semi-supervised medical image segmentation. Code is available at \textcolor{cyan}{https://github.com/maxwell0027/LeFeD}
We present a general and simple text to video model based on Transformer. Since both text and video are sequential data, we encode both texts and images into the same hidden space, which are further fed into Transformer to capture the temporal consistency and then decoder to generate either text or images. Considering the image signal may become weak in the long sequence, we introduce the U-Net to reconstruct image from its noised version. Specifically, we increase the noise level to the original image in the long sequence, then use the $down$ module from U-Net to encode noised images, which are further input to transformer to predict next clear images. We also add a constraint to promote motion between any generated image pair in the video. We use GPT2 and test our approach on UCF101 dataset and show it can generate promising videos.
Monocular 3D Semantic Scene Completion (SSC) has garnered significant attention in recent years due to its potential to predict complex semantics and geometry shapes from a single image, requiring no 3D inputs. In this paper, we identify several critical issues in current state-of-the-art methods, including the Feature Ambiguity of projected 2D features in the ray to the 3D space, the Pose Ambiguity of the 3D convolution, and the Computation Imbalance in the 3D convolution across different depth levels. To address these problems, we devise a novel Normalized Device Coordinates scene completion network (NDC-Scene) that directly extends the 2D feature map to a Normalized Device Coordinates (NDC) space, rather than to the world space directly, through progressive restoration of the dimension of depth with deconvolution operations. Experiment results demonstrate that transferring the majority of computation from the target 3D space to the proposed normalized device coordinates space benefits monocular SSC tasks. Additionally, we design a Depth-Adaptive Dual Decoder to simultaneously upsample and fuse the 2D and 3D feature maps, further improving overall performance. Our extensive experiments confirm that the proposed method consistently outperforms state-of-the-art methods on both outdoor SemanticKITTI and indoor NYUv2 datasets. Our code are available at https://github.com/Jiawei-Yao0812/NDCScene.
This work proposes a frame-wise online/streaming end-to-end neural diarization (FS-EEND) method in a frame-in-frame-out fashion. To frame-wisely detect a flexible number of speakers and extract/update their corresponding attractors, we propose to leverage a causal speaker embedding encoder and an online non-autoregressive self-attention-based attractor decoder. A look-ahead mechanism is adopted to allow leveraging some future frames for effectively detecting new speakers in real time and adaptively updating speaker attractors. The proposed method processes the audio stream frame by frame, and has a low inference latency caused by the look-ahead frames. Experiments show that, compared with the recently proposed block-wise online methods, our method FS-EEND achieves state-of-the-art diarization results, with a low inference latency and computational cost.
With the advent of AI and computer vision techniques, the quest for automated and efficient floor plan designs has gained momentum. This paper presents a novel approach using skip-connected neural networks integrated with layout graphs. The skip-connected layers capture multi-scale floor plan information, and the encoder-decoder networks with GNN facilitate pixel-level probability-based generation. Validated on the MSD dataset, our approach achieved a 93.9 mIoU score in the 1st CVAAD workshop challenge. Code and pre-trained models are publicly available at https://github.com/yuntaeJ/SkipNet-FloorPlanGe.
Learned B-frame video compression aims to adopt bi-directional motion estimation and motion compensation (MEMC) coding for middle frame reconstruction. However, previous learned approaches often directly extend neural P-frame codecs to B-frame relying on bi-directional optical-flow estimation or video frame interpolation. They suffer from inaccurate quantized motions and inefficient motion compensation. To address these issues, we propose a simple yet effective structure called Interpolation-driven B-frame Video Compression (IBVC). Our approach only involves two major operations: video frame interpolation and artifact reduction compression. IBVC introduces a bit-rate free MEMC based on interpolation, which avoids optical-flow quantization and additional compression distortions. Later, to reduce duplicate bit-rate consumption and focus on unaligned artifacts, a residual guided masking encoder is deployed to adaptively select the meaningful contexts with interpolated multi-scale dependencies. In addition, a conditional spatio-temporal decoder is proposed to eliminate location errors and artifacts instead of using MEMC coding in other methods. The experimental results on B-frame coding demonstrate that IBVC has significant improvements compared to the relevant state-of-the-art methods. Meanwhile, our approach can save bit rates compared with the random access (RA) configuration of H.266 (VTM). The code will be available at https://github.com/ruhig6/IBVC.
Although convolutional neural networks (CNNs) have been proposed to remove adverse weather conditions in single images using a single set of pre-trained weights, they fail to restore weather videos due to the absence of temporal information. Furthermore, existing methods for removing adverse weather conditions (e.g., rain, fog, and snow) from videos can only handle one type of adverse weather. In this work, we propose the first framework for restoring videos from all adverse weather conditions by developing a video adverse-weather-component suppression network (ViWS-Net). To achieve this, we first devise a weather-agnostic video transformer encoder with multiple transformer stages. Moreover, we design a long short-term temporal modeling mechanism for weather messenger to early fuse input adjacent video frames and learn weather-specific information. We further introduce a weather discriminator with gradient reversion, to maintain the weather-invariant common information and suppress the weather-specific information in pixel features, by adversarially predicting weather types. Finally, we develop a messenger-driven video transformer decoder to retrieve the residual weather-specific feature, which is spatiotemporally aggregated with hierarchical pixel features and refined to predict the clean target frame of input videos. Experimental results, on benchmark datasets and real-world weather videos, demonstrate that our ViWS-Net outperforms current state-of-the-art methods in terms of restoring videos degraded by any weather condition.
Various deep learning models have been proposed for 3D bone shape reconstruction from two orthogonal (biplanar) X-ray images. However, it is unclear how these models compare against each other since they are evaluated on different anatomy, cohort and (often privately held) datasets. Moreover, the impact of the commonly optimized image-based segmentation metrics such as dice score on the estimation of clinical parameters relevant in 2D-3D bone shape reconstruction is not well known. To move closer toward clinical translation, we propose a benchmarking framework that evaluates tasks relevant to real-world clinical scenarios, including reconstruction of fractured bones, bones with implants, robustness to population shift, and error in estimating clinical parameters. Our open-source platform provides reference implementations of 8 models (many of whose implementations were not publicly available), APIs to easily collect and preprocess 6 public datasets, and the implementation of automatic clinical parameter and landmark extraction methods. We present an extensive evaluation of 8 2D-3D models on equal footing using 6 public datasets comprising images for four different anatomies. Our results show that attention-based methods that capture global spatial relationships tend to perform better across all anatomies and datasets; performance on clinically relevant subgroups may be overestimated without disaggregated reporting; ribs are substantially more difficult to reconstruct compared to femur, hip and spine; and the dice score improvement does not always bring a corresponding improvement in the automatic estimation of clinically relevant parameters.
Masked point modeling has become a promising scheme of self-supervised pre-training for point clouds. Existing methods reconstruct either the original points or related features as the objective of pre-training. However, considering the diversity of downstream tasks, it is necessary for the model to have both low- and high-level representation modeling capabilities to capture geometric details and semantic contexts during pre-training. To this end, M$^3$CS is proposed to enable the model with the above abilities. Specifically, with masked point cloud as input, M$^3$CS introduces two decoders to predict masked representations and the original points simultaneously. While an extra decoder doubles parameters for the decoding process and may lead to overfitting, we propose siamese decoders to keep the amount of learnable parameters unchanged. Further, we propose an online codebook projecting continuous tokens into discrete ones before reconstructing masked points. In such way, we can enforce the decoder to take effect through the combinations of tokens rather than remembering each token. Comprehensive experiments show that M$^3$CS achieves superior performance at both classification and segmentation tasks, outperforming existing methods.
Biomedical literature often uses complex language and inaccessible professional terminologies. That is why simplification plays an important role in improving public health literacy. Applying Natural Language Processing (NLP) models to automate such tasks allows for quick and direct accessibility for lay readers. In this work, we investigate the ability of state-of-the-art large language models (LLMs) on the task of biomedical abstract simplification, using the publicly available dataset for plain language adaptation of biomedical abstracts (\textbf{PLABA}). The methods applied include domain fine-tuning and prompt-based learning (PBL) on: 1) Encoder-decoder models (T5, SciFive, and BART), 2) Decoder-only GPT models (GPT-3.5 and GPT-4) from OpenAI and BioGPT, and 3) Control-token mechanisms on BART-based models. We used a range of automatic evaluation metrics, including BLEU, ROUGE, SARI, and BERTscore, and also conducted human evaluations. BART-Large with Control Token (BART-L-w-CT) mechanisms reported the highest SARI score of 46.54 and T5-base reported the highest BERTscore 72.62. In human evaluation, BART-L-w-CTs achieved a better simplicity score over T5-Base (2.9 vs. 2.2), while T5-Base achieved a better meaning preservation score over BART-L-w-CTs (3.1 vs. 2.6). We also categorised the system outputs with examples, hoping this will shed some light for future research on this task. Our code, fine-tuned models, and data splits are available at \url{https://github.com/HECTA-UoM/PLABA-MU} \begin{IEEEkeywords} Large Language Models, Text Simplification, Biomedical NLP, Control Mechanisms, Health Informatics \end{IEEEkeywords}