Artistic style transfer is usually performed between two images, a style image and a content image. Recently, a model named CLIPStyler demonstrated that a natural language description of style could replace the necessity of a reference style image. They achieved this by taking advantage of the CLIP model, which can compute the similarity between a text phrase and an image. In this work, we demonstrate how combining CLIPStyler with a pre-trained, purely vision-based style transfer model can significantly reduce the inference time of CLIPStyler. We call this model FastCLIPStyler. We do a qualitative exploration of the stylised images from both models and argue that our model also has merits in terms of the visual aesthetics of the generated images. Finally, we also point out how FastCLIPStyler can be used to further extend this line of research to create a generalised text-to-style model that does not require optimisation at inference time, which both CLIPStyler and FastCLIPStyler do currently.
Multimodal emotion analysis performed better in emotion recognition depending on more comprehensive emotional clues and multimodal emotion dataset. In this paper, we developed a large multimodal emotion dataset, named "HED" dataset, to facilitate the emotion recognition task, and accordingly propose a multimodal emotion recognition method. To promote recognition accuracy, "Feature After Feature" framework was used to explore crucial emotional information from the aligned face, body and text samples. We employ various benchmarks to evaluate the "HED" dataset and compare the performance with our method. The results show that the five classification accuracy of the proposed multimodal fusion method is about 83.75%, and the performance is improved by 1.83%, 9.38%, and 21.62% respectively compared with that of individual modalities. The complementarity between each channel is effectively used to improve the performance of emotion recognition. We had also established a multimodal online emotion prediction platform, aiming to provide free emotion prediction to more users.
Table structure recognition aims to extract the logical and physical structure of unstructured table images into a machine-readable format. The latest end-to-end image-to-text approaches simultaneously predict the two structures by two decoders, where the prediction of the physical structure (the bounding boxes of the cells) is based on the representation of the logical structure. However, the previous methods struggle with imprecise bounding boxes as the logical representation lacks local visual information. To address this issue, we propose an end-to-end sequential modeling framework for table structure recognition called VAST. It contains a novel coordinate sequence decoder triggered by the representation of the non-empty cell from the logical structure decoder. In the coordinate sequence decoder, we model the bounding box coordinates as a language sequence, where the left, top, right and bottom coordinates are decoded sequentially to leverage the inter-coordinate dependency. Furthermore, we propose an auxiliary visual-alignment loss to enforce the logical representation of the non-empty cells to contain more local visual details, which helps produce better cell bounding boxes. Extensive experiments demonstrate that our proposed method can achieve state-of-the-art results in both logical and physical structure recognition. The ablation study also validates that the proposed coordinate sequence decoder and the visual-alignment loss are the keys to the success of our method.
Although large foundation models pre-trained by self-supervised learning have achieved state-of-the-art performance in many tasks including automatic speech recognition (ASR), knowledge distillation (KD) is often required in practice to transfer the knowledge learned by large teacher models into much smaller student models with affordable computation and memory costs. This paper proposes a novel two-stage KD framework to distil the knowledge from multiple speech foundation models as teachers into a single student neural transducer model for ASR. In the first stage, the student model encoder is pre-trained using the embeddings extracted from multiple teacher models. In the second stage, the student encoder is fine-tuned with the audio-text pairs based on the ASR task. Experiments on the LibriSpeech 100-hour subset show that the proposed KD framework improves the performance of both streaming and non-streaming student models when using only one teacher. The performance of the student model can be further enhanced when multiple teachers are used jointly, achieving word error rate reductions (WERRs) of 17.5% and 10.6%. Our proposed framework can be combined with other existing KD methods to achieve further improvements. Further WERRs were obtained by incorporating extra unlabelled data during encoder pre-training, leading to a total relative WERR of 55.0% on the non-streaming student model.
Despite remarkable recent advances, making object-centric learning work for complex natural scenes remains the main challenge. The recent success of adopting the transformer-based image generative model in object-centric learning suggests that having a highly expressive image generator is crucial for dealing with complex scenes. In this paper, inspired by this observation, we aim to answer the following question: can we benefit from the other pillar of modern deep generative models, i.e., the diffusion models, for object-centric learning and what are the pros and cons of such a model? To this end, we propose a new object-centric learning model, Latent Slot Diffusion (LSD). LSD can be seen from two perspectives. From the perspective of object-centric learning, it replaces the conventional slot decoders with a latent diffusion model conditioned on the object slots. Conversely, from the perspective of diffusion models, it is the first unsupervised compositional conditional diffusion model which, unlike traditional diffusion models, does not require supervised annotation such as a text description to learn to compose. In experiments on various object-centric tasks, including the FFHQ dataset for the first time in this line of research, we demonstrate that LSD significantly outperforms the state-of-the-art transformer-based decoder, particularly when the scene is more complex. We also show a superior quality in unsupervised compositional generation.
Motivated by recent advancements in text-to-image diffusion, we study erasure of specific concepts from the model's weights. While Stable Diffusion has shown promise in producing explicit or realistic artwork, it has raised concerns regarding its potential for misuse. We propose a fine-tuning method that can erase a visual concept from a pre-trained diffusion model, given only the name of the style and using negative guidance as a teacher. We benchmark our method against previous approaches that remove sexually explicit content and demonstrate its effectiveness, performing on par with Safe Latent Diffusion and censored training. To evaluate artistic style removal, we conduct experiments erasing five modern artists from the network and conduct a user study to assess the human perception of the removed styles. Unlike previous methods, our approach can remove concepts from a diffusion model permanently rather than modifying the output at the inference time, so it cannot be circumvented even if a user has access to model weights. Our code, data, and results are available at https://erasing.baulab.info/
The opaqueness of deep NLP models has motivated the development of methods for interpreting how deep models predict. Recently, work has introduced hierarchical attribution, which produces a hierarchical clustering of words, along with an attribution score for each cluster. However, existing work on hierarchical attribution all follows the connecting rule, limiting the cluster to a continuous span in the input text. We argue that the connecting rule as an additional prior may undermine the ability to reflect the model decision process faithfully. To this end, we propose to generate hierarchical explanations without the connecting rule and introduce a framework for generating hierarchical clusters. Experimental results and further analysis show the effectiveness of the proposed method in providing high-quality explanations for reflecting model predicting process.
While strides have been made in deep learning based Bengali Optical Character Recognition (OCR) in the past decade, the absence of large Document Layout Analysis (DLA) datasets has hindered the application of OCR in document transcription, e.g., transcribing historical documents and newspapers. Moreover, rule-based DLA systems that are currently being employed in practice are not robust to domain variations and out-of-distribution layouts. To this end, we present the first multidomain large Bengali Document Layout Analysis Dataset: BaDLAD. This dataset contains 33,695 human annotated document samples from six domains - i) books and magazines, ii) public domain govt. documents, iii) liberation war documents, iv) newspapers, v) historical newspapers, and vi) property deeds, with 710K polygon annotations for four unit types: text-box, paragraph, image, and table. Through preliminary experiments benchmarking the performance of existing state-of-the-art deep learning architectures for English DLA, we demonstrate the efficacy of our dataset in training deep learning based Bengali document digitization models.
A novel speech feature fusion algorithm with independent vector analysis (IVA) and parallel convolutional neural network (PCNN) is proposed for text-independent speaker recognition. Firstly, some different feature types, such as the time domain (TD) features and the frequency domain (FD) features, can be extracted from a speaker's speech, and the TD and the FD features can be considered as the linear mixtures of independent feature components (IFCs) with an unknown mixing system. To estimate the IFCs, the TD and the FD features of the speaker's speech are concatenated to build the TD and the FD feature matrix, respectively. Then, a feature tensor of the speaker's speech is obtained by paralleling the TD and the FD feature matrix. To enhance the dependence on different feature types and remove the redundancies of the same feature type, the independent vector analysis (IVA) can be used to estimate the IFC matrices of TD and FD features with the feature tensor. The IFC matrices are utilized as the input of the PCNN to extract the deep features of the TD and FD features, respectively. The deep features can be integrated to obtain the fusion feature of the speaker's speech. Finally, the fusion feature of the speaker's speech is employed as the input of a deep convolutional neural network (DCNN) classifier for speaker recognition. The experimental results show the effectiveness and performances of the proposed speaker recognition system.
This paper scrutinizes a database of over 4900 YouTube videos to characterize financial market coverage. Financial market coverage generates a large number of videos. Therefore, watching these videos to derive actionable insights could be challenging and complex. In this paper, we leverage Whisper, a speech-to-text model from OpenAI, to generate a text corpus of market coverage videos from Bloomberg and Yahoo Finance. We employ natural language processing to extract insights regarding language use from the market coverage. Moreover, we examine the prominent presence of trending topics and their evolution over time, and the impacts that some individuals and organizations have on the financial market. Our characterization highlights the dynamics of the financial market coverage and provides valuable insights reflecting broad discussions regarding recent financial events and the world economy.