Document Visual Question Answering (DocVQA) refers to the task of answering questions from document images. Existing work on DocVQA only considers single-page documents. However, in real scenarios documents are mostly composed of multiple pages that should be processed altogether. In this work we extend DocVQA to the multi-page scenario. For that, we first create a new dataset, MP-DocVQA, where questions are posed over multi-page documents instead of single pages. Second, we propose a new hierarchical method, Hi-VT5, based on the T5 architecture, that overcomes the limitations of current methods to process long multi-page documents. The proposed method is based on a hierarchical transformer architecture where the encoder summarizes the most relevant information of every page and then, the decoder takes this summarized information to generate the final answer. Through extensive experimentation, we demonstrate that our method is able, in a single stage, to answer the questions and provide the page that contains the relevant information to find the answer, which can be used as a kind of explainability measure.
Creating realistic virtual assets is a time-consuming process: it usually involves an artist designing the object, then spending a lot of effort on tweaking its appearance. Intricate details and certain effects, such as subsurface scattering, elude representation using real-time BRDFs, making it impossible to fully capture the appearance of certain objects. Inspired by the recent progress of neural rendering, we propose an approach for capturing real-world objects in everyday environments faithfully and fast. We use a novel neural representation to reconstruct volumetric effects, such as translucent object parts, and preserve photorealistic object appearance. To support real-time rendering without compromising rendering quality, our model uses a grid of features and a small MLP decoder that is transpiled into efficient shader code with interactive framerates. This leads to a seamless integration of the proposed neural assets with existing mesh environments and objects. Thanks to the use of standard shader code rendering is portable across many existing hardware and software systems.
Spatio-temporal modeling as a canonical task of multivariate time series forecasting has been a significant research topic in AI community. To address the underlying heterogeneity and non-stationarity implied in the graph streams, in this study, we propose Spatio-Temporal Meta-Graph Learning as a novel Graph Structure Learning mechanism on spatio-temporal data. Specifically, we implement this idea into Meta-Graph Convolutional Recurrent Network (MegaCRN) by plugging the Meta-Graph Learner powered by a Meta-Node Bank into GCRN encoder-decoder. We conduct a comprehensive evaluation on two benchmark datasets (METR-LA and PEMS-BAY) and a large-scale spatio-temporal dataset that contains a variaty of non-stationary phenomena. Our model outperformed the state-of-the-arts to a large degree on all three datasets (over 27% MAE and 34% RMSE). Besides, through a series of qualitative evaluations, we demonstrate that our model can explicitly disentangle locations and time slots with different patterns and be robustly adaptive to different anomalous situations. Codes and datasets are available at https://github.com/deepkashiwa20/MegaCRN.
The small object semantic segmentation task is aimed at automatically extracting key objects from high-resolution remote sensing (HRS) imagery. Compared with the large-scale coverage areas for remote sensing imagery, the key objects such as cars, ships, etc. in HRS imagery often contain only a few pixels. In this paper, to tackle this problem, the foreground activation (FA) driven small object semantic segmentation (FactSeg) framework is proposed from perspectives of structure and optimization. In the structure design, FA object representation is proposed to enhance the awareness of the weak features in small objects. The FA object representation framework is made up of a dual-branch decoder and collaborative probability (CP) loss. In the dual-branch decoder, the FA branch is designed to activate the small object features (activation), as well as suppress the largescale background, and the semantic refinement (SR) branch is designed to further distinguish small objects (refinement). The CP loss is proposed to effectively combine the activation and refinement outputs of the decoder under the CP hypothesis. During the collaboration, the weak features of the small objects are enhanced with the activation output, and the refined output can be viewed as the refinement of the binary outputs. In the optimization stage, small object mining (SOM) based network optimization is applied to automatically select effective samples, to refine the direction of the optimization, while addressing the imbalanced sample problem between the small objects and the large-scale background. The experimental results obtained with two benchmark HRS imagery segmentation datasets demonstrate that the proposed framework outperforms the state-of-the-art semantic segmentation methods, and achieves a good tradeoff between accuracy and efficiency.
In this paper, we present a novel and effective framework, named 4K-NeRF, to pursue high fidelity view synthesis on the challenging scenarios of ultra high resolutions, building on the methodology of neural radiance fields (NeRF). The rendering procedure of NeRF-based methods typically relies on a pixel-wise manner in which rays (or pixels) are treated independently on both training and inference phases, limiting its representational ability on describing subtle details, especially when lifting to a extremely high resolution. We address the issue by exploring ray correlation to enhance high-frequency details recovery. Particularly, we use the 3D-aware encoder to model geometric information effectively in a lower resolution space and recover fine details through the 3D-aware decoder, conditioned on ray features and depths estimated by the encoder. Joint training with patch-based sampling further facilitates our method incorporating the supervision from perception oriented regularization beyond pixel-wise loss. Benefiting from the use of geometry-aware local context, our method can significantly boost rendering quality on high-frequency details compared with modern NeRF methods, and achieve the state-of-the-art visual quality on 4K ultra-high-resolution scenarios. Code Available at \url{https://github.com/frozoul/4K-NeRF}
Even for us, it can be challenging to comprehend the meaning of songs. As part of this project, we explore the process of generating the meaning of songs. Despite the widespread use of text-to-text models, few attempts have been made to achieve a similar objective. Songs are primarily studied in the context of sentiment analysis. This involves identifying opinions and emotions in texts, evaluating them as positive or negative, and utilizing these evaluations to make music recommendations. In this paper, we present a generative model that offers implicit meanings for several lines of a song. Our model uses a decoder Transformer architecture GPT-2, where the input is the lyrics of a song. Furthermore, we compared the performance of this architecture with that of the encoder-decoder Transformer architecture of the T5 model. We also examined the effect of different prompt types with the option of appending additional information, such as the name of the artist and the title of the song. Moreover, we tested different decoding methods with different training parameters and evaluated our results using ROUGE. In order to build our dataset, we utilized the 'Genious' API, which allowed us to acquire the lyrics of songs and their explanations, as well as their rich metadata.
Variational autoencoders (VAEs) are one class of generative probabilistic latent-variable models designed for inference based on known data. We develop three variations on VAEs by introducing a second parameterized encoder/decoder pair and, for one variation, an additional fixed encoder. The parameters of the encoders/decoders are to be learned with a neural network. The fixed encoder is obtained by probabilistic-PCA. The variations are compared to the Evidence Lower Bound (ELBO) approximation to the original VAE. One variation leads to an Evidence Upper Bound (EUBO) that can be used in conjunction with the original ELBO to interrogate the convergence of the VAE.
In this work, we present a novel framework built to simplify 3D asset generation for amateur users. To enable interactive generation, our method supports a variety of input modalities that can be easily provided by a human, including images, text, partially observed shapes and combinations of these, further allowing to adjust the strength of each input. At the core of our approach is an encoder-decoder, compressing 3D shapes into a compact latent representation, upon which a diffusion model is learned. To enable a variety of multi-modal inputs, we employ task-specific encoders with dropout followed by a cross-attention mechanism. Due to its flexibility, our model naturally supports a variety of tasks, outperforming prior works on shape completion, image-based 3D reconstruction, and text-to-3D. Most interestingly, our model can combine all these tasks into one swiss-army-knife tool, enabling the user to perform shape generation using incomplete shapes, images, and textual descriptions at the same time, providing the relative weights for each input and facilitating interactivity. Despite our approach being shape-only, we further show an efficient method to texture the generated shape using large-scale text-to-image models.
'Actions' play a vital role in how humans interact with the world. Thus, autonomous agents that would assist us in everyday tasks also require the capability to perform 'Reasoning about Actions & Change' (RAC). This has been an important research direction in Artificial Intelligence (AI) in general, but the study of RAC with visual and linguistic inputs is relatively recent. The CLEVR_HYP (Sampat et. al., 2021) is one such testbed for hypothetical vision-language reasoning with actions as the key focus. In this work, we propose a novel learning strategy that can improve reasoning about the effects of actions. We implement an encoder-decoder architecture to learn the representation of actions as vectors. We combine the aforementioned encoder-decoder architecture with existing modality parsers and a scene graph question answering model to evaluate our proposed system on the CLEVR_HYP dataset. We conduct thorough experiments to demonstrate the effectiveness of our proposed approach and discuss its advantages over previous baselines in terms of performance, data efficiency, and generalization capability.
Domain adaptation is one of the prominent strategies for handling both domain shift, that is widely encountered in large-scale land use/land cover map calculation, and the scarcity of pixel-level ground truth that is crucial for supervised semantic segmentation. Studies focusing on adversarial domain adaptation via re-styling source domain samples, commonly through generative adversarial networks, have reported varying levels of success, yet they suffer from semantic inconsistencies, visual corruptions, and often require a large number of target domain samples. In this letter, we propose a new unsupervised domain adaptation method for the semantic segmentation of very high resolution images, that i) leads to semantically consistent and noise-free images, ii) operates with a single target domain sample (i.e. one-shot) and iii) at a fraction of the number of parameters required from state-of-the-art methods. More specifically an image-to-image translation paradigm is proposed, based on an encoder-decoder principle where latent content representations are mixed across domains, and a perceptual network module and loss function is further introduced to enforce semantic consistency. Cross-city comparative experiments have shown that the proposed method outperforms state-of-the-art domain adaptation methods. Our source code will be available at \url{https://github.com/Sarmadfismael/LRM_I2I}.
Current audio-visual separation methods share a standard architecture design where an audio encoder-decoder network is fused with visual encoding features at the encoder bottleneck. This design confounds the learning of multi-modal feature encoding with robust sound decoding for audio separation. To generalize to a new instrument: one must finetune the entire visual and audio network for all musical instruments. We re-formulate visual-sound separation task and propose Instrument as Query (iQuery) with a flexible query expansion mechanism. Our approach ensures cross-modal consistency and cross-instrument disentanglement. We utilize "visually named" queries to initiate the learning of audio queries and use cross-modal attention to remove potential sound source interference at the estimated waveforms. To generalize to a new instrument or event class, drawing inspiration from the text-prompt design, we insert an additional query as an audio prompt while freezing the attention mechanism. Experimental results on three benchmarks demonstrate that our iQuery improves audio-visual sound source separation performance.
We present a novel neural model for modern poetry generation in French. The model consists of two pretrained neural models that are fine-tuned for the poem generation task. The encoder of the model is a RoBERTa based one while the decoder is based on GPT-2. This way the model can benefit from the superior natural language understanding performance of RoBERTa and the good natural language generation performance of GPT-2. Our evaluation shows that the model can create French poetry successfully. On a 5 point scale, the lowest score of 3.57 was given by human judges to typicality and emotionality of the output poetry while the best score of 3.79 was given to understandability.
Vector-Quantized (VQ-based) generative models usually consist of two basic components, i.e., VQ tokenizers and generative transformers. Prior research focuses on improving the reconstruction fidelity of VQ tokenizers but rarely examines how the improvement in reconstruction affects the generation ability of generative transformers. In this paper, we surprisingly find that improving the reconstruction fidelity of VQ tokenizers does not necessarily improve the generation. Instead, learning to compress semantic features within VQ tokenizers significantly improves generative transformers' ability to capture textures and structures. We thus highlight two competing objectives of VQ tokenizers for image synthesis: semantic compression and details preservation. Different from previous work that only pursues better details preservation, we propose Semantic-Quantized GAN (SeQ-GAN) with two learning phases to balance the two objectives. In the first phase, we propose a semantic-enhanced perceptual loss for better semantic compression. In the second phase, we fix the encoder and codebook, but enhance and finetune the decoder to achieve better details preservation. The proposed SeQ-GAN greatly improves VQ-based generative models and surpasses the GAN and Diffusion Models on both unconditional and conditional image generation. Our SeQ-GAN (364M) achieves Frechet Inception Distance (FID) of 6.25 and Inception Score (IS) of 140.9 on 256x256 ImageNet generation, a remarkable improvement over VIT-VQGAN (714M), which obtains 11.2 FID and 97.2 IS.
Recent advances on text-to-image generation have witnessed the rise of diffusion models which act as powerful generative models. Nevertheless, it is not trivial to exploit such latent variable models to capture the dependency among discrete words and meanwhile pursue complex visual-language alignment in image captioning. In this paper, we break the deeply rooted conventions in learning Transformer-based encoder-decoder, and propose a new diffusion model based paradigm tailored for image captioning, namely Semantic-Conditional Diffusion Networks (SCD-Net). Technically, for each input image, we first search the semantically relevant sentences via cross-modal retrieval model to convey the comprehensive semantic information. The rich semantics are further regarded as semantic prior to trigger the learning of Diffusion Transformer, which produces the output sentence in a diffusion process. In SCD-Net, multiple Diffusion Transformer structures are stacked to progressively strengthen the output sentence with better visional-language alignment and linguistical coherence in a cascaded manner. Furthermore, to stabilize the diffusion process, a new self-critical sequence training strategy is designed to guide the learning of SCD-Net with the knowledge of a standard autoregressive Transformer model. Extensive experiments on COCO dataset demonstrate the promising potential of using diffusion models in the challenging image captioning task. Source code is available at \url{https://github.com/YehLi/xmodaler/tree/master/configs/image_caption/scdnet}.
Despite the tremendous progress of Masked Autoencoders (MAE) in developing vision tasks such as image and video, exploring MAE in large-scale 3D point clouds remains challenging due to the inherent irregularity. In contrast to previous 3D MAE frameworks, which either design a complex decoder to infer masked information from maintained regions or adopt sophisticated masking strategies, we instead propose a much simpler paradigm. The core idea is to apply a \textbf{G}enerative \textbf{D}ecoder for MAE (GD-MAE) to automatically merges the surrounding context to restore the masked geometric knowledge in a hierarchical fusion manner. In doing so, our approach is free from introducing the heuristic design of decoders and enjoys the flexibility of exploring various masking strategies. The corresponding part costs less than \textbf{12\%} latency compared with conventional methods, while achieving better performance. We demonstrate the efficacy of the proposed method on several large-scale benchmarks: Waymo, KITTI, and ONCE. Consistent improvement on downstream detection tasks illustrates strong robustness and generalization capability. Not only our method reveals state-of-the-art results, but remarkably, we achieve comparable accuracy even with \textbf{20\%} of the labeled data on the Waymo dataset. Code will be released at https://github.com/Nightmare-n/GD-MAE.
We introduce M-VADER: a diffusion model (DM) for image generation where the output can be specified using arbitrary combinations of images and text. We show how M-VADER enables the generation of images specified using combinations of image and text, and combinations of multiple images. Previously, a number of successful DM image generation algorithms have been introduced that make it possible to specify the output image using a text prompt. Inspired by the success of those models, and led by the notion that language was already developed to describe the elements of visual contexts that humans find most important, we introduce an embedding model closely related to a vision-language model. Specifically, we introduce the embedding model S-MAGMA: a 13 billion parameter multimodal decoder combining components from an autoregressive vision-language model MAGMA and biases finetuned for semantic search.
Controlled text generation is a very important task in the arena of natural language processing due to its promising applications. In order to achieve this task we mainly introduce the novel soft prompt tuning method of using soft prompts at both encoder and decoder levels together in a T5 model and investigate the performance as the behaviour of an additional soft prompt related to the decoder of a T5 model in controlled text generation remained unexplored. Then we also investigate the feasibility of steering the output of this extended soft prompted T5 model at decoder level and finally analyse the utility of generated text to be used in AI related tasks such as training AI models with an interpretability analysis of the classifier trained with synthetic text, as there is a lack of proper analysis of methodologies in generating properly labelled data to be utilized in AI tasks. Through the performed in-depth intrinsic and extrinsic evaluations of this generation model along with the artificially generated data, we found that this model produced better results compared to the T5 model with a single soft prompt at encoder level and the sentiment classifier trained using this artificially generated data can produce comparable classification results to the results of a classifier trained with real labelled data and also the classifier decision is interpretable with respect to the input text content.
In this work, we present a new computer vision task named video object of interest segmentation (VOIS). Given a video and a target image of interest, our objective is to simultaneously segment and track all objects in the video that are relevant to the target image. This problem combines the traditional video object segmentation task with an additional image indicating the content that users are concerned with. Since no existing dataset is perfectly suitable for this new task, we specifically construct a large-scale dataset called LiveVideos, which contains 2418 pairs of target images and live videos with instance-level annotations. In addition, we propose a transformer-based method for this task. We revisit Swin Transformer and design a dual-path structure to fuse video and image features. Then, a transformer decoder is employed to generate object proposals for segmentation and tracking from the fused features. Extensive experiments on LiveVideos dataset show the superiority of our proposed method.
Event cameras, offering high temporal resolutions and high dynamic ranges, have brought a new perspective to address common challenges (e.g., motion blur and low light) in monocular depth estimation. However, how to effectively exploit the sparse spatial information and rich temporal cues from asynchronous events remains a challenging endeavor. To this end, we propose a novel event-based monocular depth estimator with recurrent transformers, namely EReFormer, which is the first pure transformer with a recursive mechanism to process continuous event streams. Technically, for spatial modeling, a novel transformer-based encoder-decoder with a spatial transformer fusion module is presented, having better global context information modeling capabilities than CNN-based methods. For temporal modeling, we design a gate recurrent vision transformer unit that introduces a recursive mechanism into transformers, improving temporal modeling capabilities while alleviating the expensive GPU memory cost. The experimental results show that our EReFormer outperforms state-of-the-art methods by a margin on both synthetic and real-world datasets. We hope that our work will attract further research to develop stunning transformers in the event-based vision community. Our open-source code can be found in the supplemental material.
Multivariate time series forecasting (MTSF) is a fundamental problem in numerous real-world applications. Recently, Transformer has become the de facto solution for MTSF, especially for the long-term cases. However, except for the one forward operation, the basic configurations in existing MTSF Transformer architectures were barely carefully verified. In this study, we point out that the current tokenization strategy in MTSF Transformer architectures ignores the token uniformity inductive bias of Transformers. Therefore, the vanilla MTSF transformer struggles to capture details in time series and presents inferior performance. Based on this observation, we make a series of evolution on the basic architecture of the vanilla MTSF transformer. We vary the flawed tokenization strategy, along with the decoder structure and embeddings. Surprisingly, the evolved simple transformer architecture is highly effective, which successfully avoids the over-smoothing phenomena in the vanilla MTSF transformer, achieves a more detailed and accurate prediction, and even substantially outperforms the state-of-the-art Transformers that are well-designed for MTSF.