PyTorchKR Papers

최신 논문

총 9,339편
태그·기간 등 상세 필터

카테고리 탐색

로그인 후 이 태그 구독

arXiv:2304.13949

UCF: Uncovering Common Features for Generalizable Deepfake Detection

Deepfake detection remains a challenging task due to the difficulty of generalizing to new types of forgeries. This problem primarily stems from the overfitting of existing detection methods to forgery-irrelevant features and method-specific patterns. The latter has been rarely studied and not well addressed by previous works. This paper presents a novel approach to address the two types of overfitting issues by uncovering common forgery features. Specifically, we first propose a disentanglement framework that decomposes image information into three distinct components: forgery-irrelevant, method-specific forgery, and common forgery features. To ensure the decoupling of method-specific and common forgery features, a multi-task learning strategy is employed, including a multi-class classification that predicts the category of the forgery method and a binary classification that distinguishes the real from the fake. Additionally, a conditional decoder is designed to utilize forgery features as a condition along with forgery-irrelevant features to generate reconstructed images. Furthermore, a contrastive regularization technique is proposed to encourage the disentanglement of the common and specific forgery features. Ultimately, we only utilize the common forgery features for the purpose of generalizable deepfake detection. Extensive evaluations demonstrate that our framework can perform superior generalization than current state-of-the-art methods.

arXiv:2304.13785공식 코드 · Papers with Code

Customized Segment Anything Model for Medical Image Segmentation

We propose SAMed, a general solution for medical image segmentation. Different from the previous methods, SAMed is built upon the large-scale image segmentation model, Segment Anything Model (SAM), to explore the new research paradigm of customizing large-scale models for medical image segmentation. SAMed applies the low-rank-based (LoRA) finetuning strategy to the SAM image encoder and finetunes it together with the prompt encoder and the mask decoder on labeled medical image segmentation datasets. We also observe the warmup finetuning strategy and the AdamW optimizer lead SAMed to successful convergence and lower loss. Different from SAM, SAMed could perform semantic segmentation on medical images. Our trained SAMed model achieves 81.88 DSC and 20.64 HD on the Synapse multi-organ segmentation dataset, which is on par with the state-of-the-art methods. We conduct extensive experiments to validate the effectiveness of our design. Since SAMed only updates a small fraction of the SAM parameters, its deployment cost and storage cost are quite marginal in practical usage. The code of SAMed is available at https://github.com/hitachinsk/SAMed.

arXiv:2304.14394공식 코드 · Papers with Code

Unified Sequence-to-Sequence Learning for Single- and Multi-Modal Visual Object Tracking

In this paper, we introduce a new sequence-to-sequence learning framework for RGB-based and multi-modal object tracking. First, we present SeqTrack for RGB-based tracking. It casts visual tracking as a sequence generation task, forecasting object bounding boxes in an autoregressive manner. This differs from previous trackers, which depend on the design of intricate head networks, such as classification and regression heads. SeqTrack employs a basic encoder-decoder transformer architecture. The encoder utilizes a bidirectional transformer for feature extraction, while the decoder generates bounding box sequences autoregressively using a causal transformer. The loss function is a plain cross-entropy. Second, we introduce SeqTrackv2, a unified sequence-to-sequence framework for multi-modal tracking tasks. Expanding upon SeqTrack, SeqTrackv2 integrates a unified interface for auxiliary modalities and a set of task-prompt tokens to specify the task. This enables it to manage multi-modal tracking tasks using a unified model and parameter set. This sequence learning paradigm not only simplifies the tracking framework, but also showcases superior performance across 14 challenging benchmarks spanning five single- and multi-modal tracking tasks. The code and models are available at https://github.com/chenxin-dlut/SeqTrackv2.

arXiv:2304.13854

Understand the Dynamic World: An End-to-End Knowledge Informed Framework for Open Domain Entity State Tracking

Open domain entity state tracking aims to predict reasonable state changes of entities (i.e., [attribute] of [entity] was [before_state] and [after_state] afterwards) given the action descriptions. It's important to many reasoning tasks to support human everyday activities. However, it's challenging as the model needs to predict an arbitrary number of entity state changes caused by the action while most of the entities are implicitly relevant to the actions and their attributes as well as states are from open vocabularies. To tackle these challenges, we propose a novel end-to-end Knowledge Informed framework for open domain Entity State Tracking, namely KIEST, which explicitly retrieves the relevant entities and attributes from external knowledge graph (i.e., ConceptNet) and incorporates them to autoregressively generate all the entity state changes with a novel dynamic knowledge grained encoder-decoder framework. To enforce the logical coherence among the predicted entities, attributes, and states, we design a new constraint decoding strategy and employ a coherence reward to improve the decoding process. Experimental results show that our proposed KIEST framework significantly outperforms the strong baselines on the public benchmark dataset OpenPI.

arXiv:2304.13292

Zero-Shot Slot and Intent Detection in Low-Resource Languages

Intent detection and slot filling are critical tasks in spoken and natural language understanding for task-oriented dialog systems. In this work we describe our participation in the slot and intent detection for low-resource language varieties (SID4LR; Aepli et al. (2023)). We investigate the slot and intent detection (SID) tasks using a wide range of models and settings. Given the recent success of multitask-prompted finetuning of large language models, we also test the generalization capability of the recent encoder-decoder model mT0 (Muennighoff et al., 2022) on new tasks (i.e., SID) in languages they have never intentionally seen. We show that our best model outperforms the baseline by a large margin (up to +30 F1 points) in both SID tasks

arXiv:2304.13178

Robust Non-Linear Feedback Coding via Power-Constrained Deep Learning

The design of codes for feedback-enabled communications has been a long-standing open problem. Recent research on non-linear, deep learning-based coding schemes have demonstrated significant improvements in communication reliability over linear codes, but are still vulnerable to the presence of forward and feedback noise over the channel. In this paper, we develop a new family of non-linear feedback codes that greatly enhance robustness to channel noise. Our autoencoder-based architecture is designed to learn codes based on consecutive blocks of bits, which obtains de-noising advantages over bit-by-bit processing to help overcome the physical separation between the encoder and decoder over a noisy channel. Moreover, we develop a power control layer at the encoder to explicitly incorporate hardware constraints into the learning optimization, and prove that the resulting average power constraint is satisfied asymptotically. Numerical experiments demonstrate that our scheme outperforms state-of-the-art feedback codes by wide margins over practical forward and feedback noise regimes, and provide information-theoretic insights on the behavior of our non-linear codes. Moreover, we observe that, in a long blocklength regime, canonical error correction codes are still preferable to feedback codes when the feedback noise becomes high.

arXiv:2304.13649공식 코드 · Papers with Code

A Symmetric Dual Encoding Dense Retrieval Framework for Knowledge-Intensive Visual Question Answering

Knowledge-Intensive Visual Question Answering (KI-VQA) refers to answering a question about an image whose answer does not lie in the image. This paper presents a new pipeline for KI-VQA tasks, consisting of a retriever and a reader. First, we introduce DEDR, a symmetric dual encoding dense retrieval framework in which documents and queries are encoded into a shared embedding space using uni-modal (textual) and multi-modal encoders. We introduce an iterative knowledge distillation approach that bridges the gap between the representation spaces in these two encoders. Extensive evaluation on two well-established KI-VQA datasets, i.e., OK-VQA and FVQA, suggests that DEDR outperforms state-of-the-art baselines by 11.6% and 30.9% on OK-VQA and FVQA, respectively. Utilizing the passages retrieved by DEDR, we further introduce MM-FiD, an encoder-decoder multi-modal fusion-in-decoder model, for generating a textual answer for KI-VQA tasks. MM-FiD encodes the question, the image, and each retrieved passage separately and uses all passages jointly in its decoder. Compared to competitive baselines in the literature, this approach leads to 5.5% and 8.5% improvements in terms of question answering accuracy on OK-VQA and FVQA, respectively.

arXiv:2304.13273공식 코드 · Papers with Code

From Association to Generation: Text-only Captioning by Unsupervised Cross-modal Mapping

With the development of Vision-Language Pre-training Models (VLPMs) represented by CLIP and ALIGN, significant breakthroughs have been achieved for association-based visual tasks such as image classification and image-text retrieval by the zero-shot capability of CLIP without fine-tuning. However, CLIP is hard to apply to generation-based tasks. This is due to the lack of decoder architecture and pre-training tasks for generation. Although previous works have created generation capacity for CLIP through additional language models, a modality gap between the CLIP representations of different modalities and the inability of CLIP to model the offset of this gap, which fails the concept to transfer across modalities. To solve the problem, we try to map images/videos to the language modality and generate captions from the language modality. In this paper, we propose the K-nearest-neighbor Cross-modality Mapping (Knight), a zero-shot method from association to generation. With text-only unsupervised training, Knight achieves State-of-the-Art performance in zero-shot methods for image captioning and video captioning. Our code is available at https://github.com/junyangwang0410/Knight.

arXiv:2304.13265

StepFormer: Self-supervised Step Discovery and Localization in Instructional Videos

Instructional videos are an important resource to learn procedural tasks from human demonstrations. However, the instruction steps in such videos are typically short and sparse, with most of the video being irrelevant to the procedure. This motivates the need to temporally localize the instruction steps in such videos, i.e. the task called key-step localization. Traditional methods for key-step localization require video-level human annotations and thus do not scale to large datasets. In this work, we tackle the problem with no human supervision and introduce StepFormer, a self-supervised model that discovers and localizes instruction steps in a video. StepFormer is a transformer decoder that attends to the video with learnable queries, and produces a sequence of slots capturing the key-steps in the video. We train our system on a large dataset of instructional videos, using their automatically-generated subtitles as the only source of supervision. In particular, we supervise our system with a sequence of text narrations using an order-aware loss function that filters out irrelevant phrases. We show that our model outperforms all previous unsupervised and weakly-supervised approaches on step detection and localization by a large margin on three challenging benchmarks. Moreover, our model demonstrates an emergent property to solve zero-shot multi-step localization and outperforms all relevant baselines at this task.

arXiv:2304.11509

Co-GRU Enhanced End-to-End Design for Long-haul Coherent Transmission Systems

In recent years, the end-to-end (E2E) scheme based on deep learning (DL) has been proposed as a potential scheme to jointly optimize the encoder and the decoder parameters of the optical communication system. Compared with conventional deep neural network (DNN) adopted in E2E design, center-oriented Gated Recurrent Unit (Co-GRU) network has the ability to learn and compensate for inter-symbol interference (ISI) with low computation cost while satisfying the gradient backpropagation (BP) condition. In this paper, the Co-GRU structure is adopted for both channel modeling and decoder implementation in E2E design for long-haul coherent wavelength division multiplexing (WDM) transmission systems, which can enhance the performance of general mutual information (GMI) and Q2-factor. For the E2E system with Co-GRU based decoder, the gain of GMI and Q2-factor are respectively improved 0.2 bits/sym and 0.48dB, compared to that of the conventional QAM system, for a 5-channel dual-polarization coherent system transmitting over 960km standard single mode fiber (SSMF). This work paves the way for further study of the application of the Co-GRU structure for both the data-driven channel modeling and the decoder performance improvement in E2E design.

논문 ID: document/1010804

SDSC-UNet: Dual Skip Connection ViT-based U-shaped Model for Building Extraction

Benefiting from effective global information interaction, vision-transformers (ViTs) have been widely used in the building extraction task. However, buildings in remote sensing (RS) images usually differ greatly in size. Mainstream ViT-based segmentation models for RS images are based on Swin Transformer, which lacks multi-scale information inside the ViT block. In addition, they only connect the output of the entire ViT encoder block to the decoder, which ignore the similarity information of the attention maps inside the ViT encoder block, and are unable to provide better global dependencies for the decoder. To solve above problems, we introduce a novel Shunted Transformer, which enables the model to capture multi-scale information internally while fully establishing global dependencies, to build a pure ViT-based U-shaped model for building extraction. Furthermore, unlike the previous single-skip-connection structure of U-shaped methods, we build a novel dual skip connection structure inside the model. It simultaneously transmits the attention maps inside the ViT encoder block and its entire output to the decoder, thereby fully mining the information of the ViT encoder block and providing better global information guidance for the decoder. Thus, our model is named Shunted Dual Skip Connection UNet (SDSC-UNet). We also design a feature fusion module called Dual Skip Upsample Fusion Module (DSUFM) to aggregate the information. Our model has yields state-of-the-art (SOTA) performance (83.02%IoU) on the Inria Aerial Image Labeling Dataset. Code will be available at: https://github.com/stdcoutzrh/BuildingExtraction.

arXiv:2304.10869

Non-autoregressive End-to-end Approaches for Joint Automatic Speech Recognition and Spoken Language Understanding

This paper presents the use of non-autoregressive (NAR) approaches for joint automatic speech recognition (ASR) and spoken language understanding (SLU) tasks. The proposed NAR systems employ a Conformer encoder that applies connectionist temporal classification (CTC) to transcribe the speech utterance into raw ASR hypotheses, which are further refined with a bidirectional encoder representations from Transformers (BERT)-like decoder. In the meantime, the intent and slot labels of the utterance are predicted simultaneously using the same decoder. Both Mask-CTC and self-conditioned CTC (SC-CTC) approaches are explored for this study. Experiments conducted on the SLURP dataset show that the proposed SC-Mask-CTC NAR system achieves 3.7% and 3.2% absolute gains in SLU metrics and a competitive level of ASR accuracy, when compared to a Conformer-Transformer based autoregressive (AR) model. Additionally, the NAR systems achieve 6x faster decoding speed than the AR baseline.

arXiv:2304.12770

Controlling Posterior Collapse by an Inverse Lipschitz Constraint on the Decoder Network

Variational autoencoders (VAEs) are one of the deep generative models that have experienced enormous success over the past decades. However, in practice, they suffer from a problem called posterior collapse, which occurs when the encoder coincides, or collapses, with the prior taking no information from the latent structure of the input data into consideration. In this work, we introduce an inverse Lipschitz neural network into the decoder and, based on this architecture, provide a new method that can control in a simple and clear manner the degree of posterior collapse for a wide range of VAE models equipped with a concrete theoretical guarantee. We also illustrate the effectiveness of our method through several numerical experiments.

arXiv:2304.12502

Causal Semantic Communication for Digital Twins: A Generalizable Imitation Learning Approach

A digital twin (DT) leverages a virtual representation of the physical world, along with communication (e.g., 6G), computing (e.g., edge computing), and artificial intelligence (AI) technologies to enable many connected intelligence services. In order to handle the large amounts of network data based on digital twins (DTs), wireless systems can exploit the paradigm of semantic communication (SC) for facilitating informed decision-making under strict communication constraints by utilizing AI techniques such as causal reasoning. In this paper, a novel framework called causal semantic communication (CSC) is proposed for DT-based wireless systems. The CSC system is posed as an imitation learning (IL) problem, where the transmitter, with access to optimal network control policies using a DT, teaches the receiver using SC over a bandwidth limited wireless channel how to improve its knowledge to perform optimal control actions. The causal structure in the source data is extracted using novel approaches from the framework of deep end-to-end causal inference, thereby enabling the creation of a semantic representation that is causally invariant, which in turn helps generalize the learned knowledge of the system to unseen scenarios. The CSC decoder at the receiver is designed to extract and estimate semantic information while ensuring high semantic reliability. The receiver control policies, semantic decoder, and causal inference are formulated as a bi-level optimization problem within a variational inference framework. This problem is solved using a novel concept called network state models, inspired from world models in generative AI, that faithfully represents the environment dynamics leading to data generation. Simulation results demonstrate that the proposed CSC system outperforms state-of-the-art SC systems by achieving better semantic reliability and reduced semantic representation.

arXiv:2304.12776

State Spaces Aren't Enough: Machine Translation Needs Attention

Structured State Spaces for Sequences (S4) is a recently proposed sequence model with successful applications in various tasks, e.g. vision, language modeling, and audio. Thanks to its mathematical formulation, it compresses its input to a single hidden state, and is able to capture long range dependencies while avoiding the need for an attention mechanism. In this work, we apply S4 to Machine Translation (MT), and evaluate several encoder-decoder variants on WMT'14 and WMT'16. In contrast with the success in language modeling, we find that S4 lags behind the Transformer by approximately 4 BLEU points, and that it counter-intuitively struggles with long sentences. Finally, we show that this gap is caused by S4's inability to summarize the full source sentence in a single hidden state, and show that we can close the gap by introducing an attention mechanism.

arXiv:2304.12532

SEA: A Spatially Explicit Architecture for Multi-Agent Reinforcement Learning

Spatial information is essential in various fields. How to explicitly model according to the spatial location of agents is also very important for the multi-agent problem, especially when the number of agents is changing and the scale is enormous. Inspired by the point cloud task in computer vision, we propose a spatial information extraction structure for multi-agent reinforcement learning in this paper. Agents can effectively share the neighborhood and global information through a spatially encoder-decoder structure. Our method follows the centralized training with decentralized execution (CTDE) paradigm. In addition, our structure can be applied to various existing mainstream reinforcement learning algorithms with minor modifications and can deal with the problem with a variable number of agents. The experiments in several multi-agent scenarios show that the existing methods can get convincing results by adding our spatially explicit architecture.

arXiv:2304.11379공식 코드 · Papers with Code

LiDAR2Map: In Defense of LiDAR-Based Semantic Map Construction Using Online Camera Distillation

Semantic map construction under bird's-eye view (BEV) plays an essential role in autonomous driving. In contrast to camera image, LiDAR provides the accurate 3D observations to project the captured 3D features onto BEV space inherently. However, the vanilla LiDAR-based BEV feature often contains many indefinite noises, where the spatial features have little texture and semantic cues. In this paper, we propose an effective LiDAR-based method to build semantic map. Specifically, we introduce a BEV feature pyramid decoder that learns the robust multi-scale BEV features for semantic map construction, which greatly boosts the accuracy of the LiDAR-based method. To mitigate the defects caused by lacking semantic cues in LiDAR data, we present an online Camera-to-LiDAR distillation scheme to facilitate the semantic learning from image to point cloud. Our distillation scheme consists of feature-level and logit-level distillation to absorb the semantic information from camera in BEV. The experimental results on challenging nuScenes dataset demonstrate the efficacy of our proposed LiDAR2Map on semantic map construction, which significantly outperforms the previous LiDAR-based methods over 27.9% mIoU and even performs better than the state-of-the-art camera-based approaches. Source code is available at: https://github.com/songw-zju/LiDAR2Map.

arXiv:2304.11330

Self-supervised Learning by View Synthesis

We present view-synthesis autoencoders (VSA) in this paper, which is a self-supervised learning framework designed for vision transformers. Different from traditional 2D pretraining methods, VSA can be pre-trained with multi-view data. In each iteration, the input to VSA is one view (or multiple views) of a 3D object and the output is a synthesized image in another target pose. The decoder of VSA has several cross-attention blocks, which use the source view as value, source pose as key, and target pose as query. They achieve cross-attention to synthesize the target view. This simple approach realizes large-angle view synthesis and learns spatial invariant representation, where the latter is decent initialization for transformers on downstream tasks, such as 3D classification on ModelNet40, ShapeNet Core55, and ScanObjectNN. VSA outperforms existing methods significantly for linear probing and is competitive for fine-tuning. The code will be made publicly available.

arXiv:2304.11546

CANet: Curved Guide Line Network with Adaptive Decoder for Lane Detection

Lane detection is challenging due to the complicated on road scenarios and line deformation from different camera perspectives. Lots of solutions were proposed, but can not deal with corner lanes well. To address this problem, this paper proposes a new top-down deep learning lane detection approach, CANET. A lane instance is first responded by the heat-map on the U-shaped curved guide line at global semantic level, thus the corresponding features of each lane are aggregated at the response point. Then CANET obtains the heat-map response of the entire lane through conditional convolution, and finally decodes the point set to describe lanes via adaptive decoder. The experimental results show that CANET reaches SOTA in different metrics. Our code will be released soon.

arXiv:2304.11450공식 코드 · Papers with Code

Dilated-UNet: A Fast and Accurate Medical Image Segmentation Approach using a Dilated Transformer and U-Net Architecture

Medical image segmentation is crucial for the development of computer-aided diagnostic and therapeutic systems, but still faces numerous difficulties. In recent years, the commonly used encoder-decoder architecture based on CNNs has been applied effectively in medical image segmentation, but has limitations in terms of learning global context and spatial relationships. Some researchers have attempted to incorporate transformers into both the decoder and encoder components, with promising results, but this approach still requires further improvement due to its high computational complexity. This paper introduces Dilated-UNet, which combines a Dilated Transformer block with the U-Net architecture for accurate and fast medical image segmentation. Image patches are transformed into tokens and fed into the U-shaped encoder-decoder architecture, with skip-connections for local-global semantic feature learning. The encoder uses a hierarchical Dilated Transformer with a combination of Neighborhood Attention and Dilated Neighborhood Attention Transformer to extract local and sparse global attention. The results of our experiments show that Dilated-UNet outperforms other models on several challenging medical image segmentation datasets, such as ISIC and Synapse.