Until recently, the field of speaker diarization was dominated by cascaded systems. Due to their limitations, mainly regarding overlapped speech and cumbersome pipelines, end-to-end models have gained great popularity lately. One of the most successful models is end-to-end neural diarization with encoder-decoder based attractors (EEND-EDA). In this work, we replace the EDA module with a Perceiver-based one and show its advantages over EEND-EDA; namely obtaining better performance on the largely studied Callhome dataset, finding the quantity of speakers in a conversation more accurately, and faster inference time. Furthermore, when exhaustively compared with other methods, our model, DiaPer, reaches remarkable performance with a very lightweight design. Besides, we perform comparisons with other works and a cascaded baseline across more than ten public wide-band datasets. Together with this publication, we release the code of DiaPer as well as models trained on public and free data.
The scarcity of labeled audio-visual datasets is a constraint for training superior audio-visual speaker diarization systems. To improve the performance of audio-visual speaker diarization, we leverage pre-trained supervised and self-supervised speech models for audio-visual speaker diarization. Specifically, we adopt supervised~(ResNet and ECAPA-TDNN) and self-supervised pre-trained models~(WavLM and HuBERT) as the speaker and audio embedding extractors in an end-to-end audio-visual speaker diarization~(AVSD) system. Then we explore the effectiveness of different frameworks, including Transformer, Conformer, and cross-attention mechanism, in the audio-visual decoder. To mitigate the degradation of performance caused by separate training, we jointly train the audio encoder, speaker encoder, and audio-visual decoder in the AVSD system. Experiments on the MISP dataset demonstrate that the proposed method achieves superior performance and obtained third place in MISP Challenge 2022.
Individual trajectories, rich in human-environment interaction information across space and time, serve as vital inputs for geospatial foundation models (GeoFMs). However, existing attempts at learning trajectory representations have overlooked the implicit spatial-temporal dependency within trajectories, failing to encode such dependency in a deep learning-friendly format. That poses a challenge in obtaining general-purpose trajectory representations. Therefore, this paper proposes a spatial-temporal joint representation learning method (ST-GraphRL) to formalize learnable spatial-temporal dependencies into trajectory representations. The proposed ST-GraphRL consists of three compositions: (i) a weighted directed spatial-temporal graph to explicitly construct mobility interactions in both space and time dimensions; (ii) a two-stage jointly encoder (i.e., decoupling and fusion), to learn entangled spatial-temporal dependencies by independently decomposing and jointly aggregating space and time information; (iii) a decoder guides ST-GraphRL to learn explicit mobility regularities by simulating the spatial-temporal distributions of trajectories. Tested on three real-world human mobility datasets, the proposed ST-GraphRL outperformed all the baseline models in predicting movement spatial-temporal distributions and preserving trajectory similarity with high spatial-temporal correlations. Analyzing spatial-temporal features presented in latent space validates that ST-GraphRL understands spatial-temporal patterns. This study may also benefit representation learnings of other geospatial data to achieve general-purpose data representations and advance GeoFMs development.
The recent emergence of Large Language Models based on the Transformer architecture has enabled dramatic advancements in the field of Natural Language Processing. However, these models have long inference latency, which limits their deployment and makes them prohibitively expensive for various real-time applications. The inference latency is further exacerbated by autoregressive generative tasks, as models need to run iteratively to generate tokens sequentially without leveraging token-level parallelization. To address this, we propose Big Little Decoder (BiLD), a framework that can improve inference efficiency and latency for a wide range of text generation applications. The BiLD framework contains two models with different sizes that collaboratively generate text. The small model runs autoregressively to generate text with a low inference cost, and the large model is only invoked occasionally to refine the small model's inaccurate predictions in a non-autoregressive manner. To coordinate the small and large models, BiLD introduces two simple yet effective policies: (1) the fallback policy that determines when to hand control over to the large model; and (2) the rollback policy that determines when the large model needs to correct the small model's inaccurate predictions. To evaluate our framework across different tasks and models, we apply BiLD to various text generation scenarios encompassing machine translation on IWSLT 2017 De-En and WMT 2014 De-En, and summarization on XSUM and CNN/DailyMail. On an NVIDIA T4 GPU, our framework achieves a speedup of up to 2.12x speedup with minimal generation quality degradation. Furthermore, our framework is fully plug-and-play and can be applied without any modifications in the training process or model architecture. Our code is open-sourced
A longstanding challenge for self-driving development is simulating dynamic driving scenarios seeded from recorded driving logs. In pursuit of this functionality, we apply tools from discrete sequence modeling to model how vehicles, pedestrians and cyclists interact in driving scenarios. Using a simple data-driven tokenization scheme, we discretize trajectories to centimeter-level resolution using a small vocabulary. We then model the multi-agent sequence of discrete motion tokens with a GPT-like encoder-decoder that is autoregressive in time and takes into account intra-timestep interaction between agents. Scenarios sampled from our model exhibit state-of-the-art realism; our model tops the Waymo Sim Agents Benchmark, surpassing prior work along the realism meta metric by 3.3% and along the interaction metric by 9.9%. We ablate our modeling choices in full autonomy and partial autonomy settings, and show that the representations learned by our model can quickly be adapted to improve performance on nuScenes. We additionally evaluate the scalability of our model with respect to parameter count and dataset size, and use density estimates from our model to quantify the saliency of context length and intra-timestep interaction for the traffic modeling task.
Watermarking of language model outputs enables statistical detection of model-generated text, which can mitigate harms and misuses of language models. Existing watermarking strategies operate by altering the decoder of an existing language model. In this paper, we ask whether language models can directly learn to generate watermarked text, which would have significant implications for the real-world deployment of watermarks. First, learned watermarks could be used to build open models that naturally generate watermarked text, enabling watermarking for open models, where users can control the decoding procedure. Second, if watermarking is used to determine the provenance of generated text, an adversary can hurt the reputation of a victim model by spoofing its watermark and generating damaging watermarked text. To investigate the learnability of watermarks, we propose watermark distillation, which trains a student model to behave like a teacher model that uses decoding-based watermarking. We test our approach on three decoding-based watermarking strategies and various hyperparameter settings, finding that models can learn to generate watermarked text with high detectability. We also find limitations to learnability, including the loss of watermarking capabilities under fine-tuning on normal text and high sample complexity when learning low-distortion watermarks.
Document image enhancement is a fundamental and important stage for attaining the best performance in any document analysis assignment because there are many degradation situations that could harm document images, making it more difficult to recognize and analyze them. In this paper, we propose \textbf{T2T-BinFormer} which is a novel document binarization encoder-decoder architecture based on a Tokens-to-token vision transformer. Each image is divided into a set of tokens with a defined length using the ViT model, which is then applied several times to model the global relationship between the tokens. However, the conventional tokenization of input data does not adequately reflect the crucial local structure between adjacent pixels of the input image, which results in low efficiency. Instead of using a simple ViT and hard splitting of images for the document image enhancement task, we employed a progressive tokenization technique to capture this local information from an image to achieve more effective results. Experiments on various DIBCO and H-DIBCO benchmarks demonstrate that the proposed model outperforms the existing CNN and ViT-based state-of-the-art methods. In this research, the primary area of examination is the application of the proposed architecture to the task of document binarization. The source code will be made available at https://github.com/RisabBiswas/T2T-BinFormer.
We present PrimDiffusion, the first diffusion-based framework for 3D human generation. Devising diffusion models for 3D human generation is difficult due to the intensive computational cost of 3D representations and the articulated topology of 3D humans. To tackle these challenges, our key insight is operating the denoising diffusion process directly on a set of volumetric primitives, which models the human body as a number of small volumes with radiance and kinematic information. This volumetric primitives representation marries the capacity of volumetric representations with the efficiency of primitive-based rendering. Our PrimDiffusion framework has three appealing properties: 1) compact and expressive parameter space for the diffusion model, 2) flexible 3D representation that incorporates human prior, and 3) decoder-free rendering for efficient novel-view and novel-pose synthesis. Extensive experiments validate that PrimDiffusion outperforms state-of-the-art methods in 3D human generation. Notably, compared to GAN-based methods, our PrimDiffusion supports real-time rendering of high-quality 3D humans at a resolution of $512\times512$ once the denoising process is done. We also demonstrate the flexibility of our framework on training-free conditional generation such as texture transfer and 3D inpainting.
In this paper, we democratise 3D content creation, enabling precise generation of 3D shapes from abstract sketches while overcoming limitations tied to drawing skills. We introduce a novel part-level modelling and alignment framework that facilitates abstraction modelling and cross-modal correspondence. Leveraging the same part-level decoder, our approach seamlessly extends to sketch modelling by establishing correspondence between CLIPasso edgemaps and projected 3D part regions, eliminating the need for a dataset pairing human sketches and 3D shapes. Additionally, our method introduces a seamless in-position editing process as a byproduct of cross-modal part-aligned modelling. Operating in a low-dimensional implicit space, our approach significantly reduces computational demands and processing time.
The ubiquity of offensive content on social media is a growing cause for concern among companies and government organizations. Recently, transformer-based models such as BERT, XLNET, and XLM-R have achieved state-of-the-art performance in detecting various forms of offensive content (e.g. hate speech, cyberbullying, and cyberaggression). However, the majority of these models are limited in their capabilities due to their encoder-only architecture, which restricts the number and types of labels in downstream tasks. Addressing these limitations, this study presents the first pre-trained model with encoder-decoder architecture for offensive language identification with text-to-text transformers (T5) trained on two large offensive language identification datasets; SOLID and CCTK. We investigate the effectiveness of combining two datasets and selecting an optimal threshold in semi-supervised instances in SOLID in the T5 retraining step. Our pre-trained T5 model outperforms other transformer-based models fine-tuned for offensive language detection, such as fBERT and HateBERT, in multiple English benchmarks. Following a similar approach, we also train the first multilingual pre-trained model for offensive language identification using mT5 and evaluate its performance on a set of six different languages (German, Hindi, Korean, Marathi, Sinhala, and Spanish). The results demonstrate that this multilingual model achieves a new state-of-the-art on all the above datasets, showing its usefulness in multilingual scenarios. Our proposed T5-based models will be made freely available to the community.
This paper presents a novel deep learning model based on the transformer architecture to predict the load-deformation behavior of large bored piles in Bangkok subsoil. The model encodes the soil profile and pile features as tokenization input, and generates the load-deformation curve as output. The model also incorporates the previous sequential data of load-deformation curve into the decoder to improve the prediction accuracy. The model also incorporates the previous sequential data of load-deformation curve into the decoder. The model shows a satisfactory accuracy and generalization ability for the load-deformation curve prediction, with a mean absolute error of 5.72% for the test data. The model could also be used for parametric analysis and design optimization of piles under different soil and pile conditions, pile cross section, pile length and type of pile.
Interactions with virtual assistants typically start with a trigger phrase followed by a command. In this work, we explore the possibility of making these interactions more natural by eliminating the need for a trigger phrase. Our goal is to determine whether a user addressed the virtual assistant based on signals obtained from the streaming audio recorded by the device microphone. We address this task by combining 1-best hypotheses and decoder signals from an automatic speech recognition system with acoustic representations from an audio encoder as input features to a large language model (LLM). In particular, we are interested in data and resource efficient systems that require only a small amount of training data and can operate in scenarios with only a single frozen LLM available on a device. For this reason, our model is trained on 80k or less examples of multimodal data using a combination of low-rank adaptation and prefix tuning. We compare the proposed system to unimodal baselines and show that the multimodal approach achieves lower equal-error-rates (EERs), while using only a fraction of the training data. We also show that low-dimensional specialized audio representations lead to lower EERs than high-dimensional general audio representations.
There has been a lot of work in question generation where different methods to provide target answers as input, have been employed. This experimentation has been mostly carried out for RNN based models. We use three different methods and their combinations for incorporating answer information and explore their effect on several automatic evaluation metrics. The methods that are used are answer prompting, using a custom product method using answer embeddings and encoder outputs, choosing sentences from the input paragraph that have answer related information, and using a separate cross-attention attention block in the decoder which attends to the answer. We observe that answer prompting without any additional modes obtains the best scores across rouge, meteor scores. Additionally, we use a custom metric to calculate how many of the generated questions have the same answer, as the answer which is used to generate them.
State-of-the-art 3D point cloud registration methods rely on labeled 3D datasets for training, which limits their practical applications in real-world scenarios and often hinders generalization to unseen scenes. Leveraging the zero-shot capabilities of foundation models offers a promising solution to these challenges. In this paper, we introduce ZeroReg, a zero-shot registration approach that utilizes 2D foundation models to predict 3D correspondences. Specifically, ZeroReg adopts an object-to-point matching strategy, starting with object localization and semantic feature extraction from multi-view images using foundation models. In the object matching stage, semantic features help identify correspondences between objects across views. However, relying solely on semantic features can lead to ambiguity, especially in scenes with multiple instances of the same category. To address this, we construct scene graphs to capture spatial relationships among objects and apply a graph matching algorithm to these graphs to accurately identify matched objects. Finally, computing fine-grained point-level correspondences within matched object regions using algorithms like SuperGlue and LoFTR achieves robust point cloud registration. Evaluations on benchmarks such as 3DMatch, 3DLoMatch, and ScanNet demonstrate ZeroReg's competitive performance, highlighting its potential to advance point-cloud registration by integrating semantic features from foundation models.
The term "Reversal Curse" refers to the scenario where auto-regressive decoder large language models (LLMs), such as ChatGPT, trained on "A is B" fail to learn "B is A," assuming that B and A are distinct and can be uniquely identified from each other, demonstrating a basic failure of logical deduction. This raises a red flag in the use of GPT models for certain general tasks such as constructing knowledge graphs, considering their adherence to this symmetric principle. In our study, we examined a bidirectional LLM, BERT, and found that it is immune to the reversal curse. Driven by ongoing efforts to construct biomedical knowledge graphs with LLMs, we also embarked on evaluating more complex but essential deductive reasoning capabilities. This process included first training encoder and decoder language models to master the intersection and union operations on two sets and then moving on to assess their capability to infer different combinations of union and intersection operations on three newly created sets. The findings showed that while both encoder and decoder language models, trained for tasks involving two sets (union/intersection), were proficient in such scenarios, they encountered difficulties when dealing with operations that included three sets (various combinations of union and intersection). Our research highlights the distinct characteristics of encoder and decoder models in simple and complex logical reasoning. In practice, the choice between BERT and GPT should be guided by the specific requirements and nature of the task at hand, leveraging their respective strengths in bidirectional context comprehension and sequence prediction.
In real life, various degradation scenarios exist that might damage document images, making it harder to recognize and analyze them, thus binarization is a fundamental and crucial step for achieving the most optimal performance in any document analysis task. We propose DocBinFormer (Document Binarization Transformer), a novel two-level vision transformer (TL-ViT) architecture based on vision transformers for effective document image binarization. The presented architecture employs a two-level transformer encoder to effectively capture both global and local feature representation from the input images. These complimentary bi-level features are exploited for efficient document image binarization, resulting in improved results for system-generated as well as handwritten document images in a comprehensive approach. With the absence of convolutional layers, the transformer encoder uses the pixel patches and sub-patches along with their positional information to operate directly on them, while the decoder generates a clean (binarized) output image from the latent representation of the patches. Instead of using a simple vision transformer block to extract information from the image patches, the proposed architecture uses two transformer blocks for greater coverage of the extracted feature space on a global and local scale. The encoded feature representation is used by the decoder block to generate the corresponding binarized output. Extensive experiments on a variety of DIBCO and H-DIBCO benchmarks show that the proposed model outperforms state-of-the-art techniques on four metrics. The source code will be made available at https://github.com/RisabBiswas/DocBinFormer.
Salient object detection (SOD) in optical remote sensing images (ORSIs) has become increasingly popular recently. Due to the characteristics of ORSIs, ORSI-SOD is full of challenges, such as multiple objects, small objects, low illuminations, and irregular shapes. To address these challenges, we propose a concise yet effective Texture-Semantic Collaboration Network (TSCNet) to explore the collaboration of texture cues and semantic cues for ORSI-SOD. Specifically, TSCNet is based on the generic encoder-decoder structure. In addition to the encoder and decoder, TSCNet includes a vital Texture-Semantic Collaboration Module (TSCM), which performs valuable feature modulation and interaction on basic features extracted from the encoder. The main idea of our TSCM is to make full use of the texture features at the lowest level and the semantic features at the highest level to achieve the expression enhancement of salient regions on features. In the TSCM, we first enhance the position of potential salient regions using semantic features. Then, we render and restore the object details using the texture features. Meanwhile, we also perceive regions of various scales, and construct interactions between different regions. Thanks to the perfect combination of TSCM and generic structure, our TSCNet can take care of both the position and details of salient objects, effectively handling various scenes. Extensive experiments on three datasets demonstrate that our TSCNet achieves competitive performance compared to 14 state-of-the-art methods. The code and results of our method are available at https://github.com/MathLee/TSCNet.
In the field of autonomous vehicles (AVs), accurately discerning commander intent and executing linguistic commands within a visual context presents a significant challenge. This paper introduces a sophisticated encoder-decoder framework, developed to address visual grounding in AVs.Our Context-Aware Visual Grounding (CAVG) model is an advanced system that integrates five core encoders-Text, Image, Context, and Cross-Modal-with a Multimodal decoder. This integration enables the CAVG model to adeptly capture contextual semantics and to learn human emotional features, augmented by state-of-the-art Large Language Models (LLMs) including GPT-4. The architecture of CAVG is reinforced by the implementation of multi-head cross-modal attention mechanisms and a Region-Specific Dynamic (RSD) layer for attention modulation. This architectural design enables the model to efficiently process and interpret a range of cross-modal inputs, yielding a comprehensive understanding of the correlation between verbal commands and corresponding visual scenes. Empirical evaluations on the Talk2Car dataset, a real-world benchmark, demonstrate that CAVG establishes new standards in prediction accuracy and operational efficiency. Notably, the model exhibits exceptional performance even with limited training data, ranging from 50% to 75% of the full dataset. This feature highlights its effectiveness and potential for deployment in practical AV applications. Moreover, CAVG has shown remarkable robustness and adaptability in challenging scenarios, including long-text command interpretation, low-light conditions, ambiguous command contexts, inclement weather conditions, and densely populated urban environments. The code for the proposed model is available at our Github.
Content moderation on a global scale must navigate a complex array of local cultural distinctions, which can hinder effective enforcement. While global policies aim for consistency and broad applicability, they often miss the subtleties of regional language interpretation, cultural beliefs, and local legislation. This work introduces a flexible framework that enhances foundation language models with cultural knowledge. Our approach involves fine-tuning encoder-decoder models on media-diet data to capture cultural nuances, and applies a continued training regime to effectively integrate these models into a content moderation pipeline. We evaluate this framework in a case study of an online podcast platform with content spanning various regions. The results show that our culturally adapted models improve the accuracy of local violation detection and offer explanations that align more closely with regional cultural norms. Our findings reinforce the need for an adaptable content moderation approach that remains flexible in response to the diverse cultural landscapes it operates in and represents a step towards a more equitable and culturally sensitive framework for content moderation, demonstrating what is achievable in this domain.
In recent years, end-to-end learnt video codecs have demonstrated their potential to compete with conventional coding algorithms in term of compression efficiency. However, most learning-based video compression models are associated with high computational complexity and latency, in particular at the decoder side, which limits their deployment in practical applications. In this paper, we present a novel model-agnostic pruning scheme based on gradient decay and adaptive layer-wise distillation. Gradient decay enhances parameter exploration during sparsification whilst preventing runaway sparsity and is superior to the standard Straight-Through Estimation. The adaptive layer-wise distillation regulates the sparse training in various stages based on the distortion of intermediate features. This stage-wise design efficiently updates parameters with minimal computational overhead. The proposed approach has been applied to three popular end-to-end learnt video codecs, FVC, DCVC, and DCVC-HEM. Results confirm that our method yields up to 65% reduction in MACs and 2x speed-up with less than 0.3dB drop in BD-PSNR. Supporting code and supplementary material can be downloaded from: https://jasminepp.github.io/lightweightdvc/