We present an effective and efficient method that explores the properties of Transformers in the frequency domain for high-quality image deblurring. Our method is motivated by the convolution theorem that the correlation or convolution of two signals in the spatial domain is equivalent to an element-wise product of them in the frequency domain. This inspires us to develop an efficient frequency domain-based self-attention solver (FSAS) to estimate the scaled dot-product attention by an element-wise product operation instead of the matrix multiplication in the spatial domain. In addition, we note that simply using the naive feed-forward network (FFN) in Transformers does not generate good deblurred results. To overcome this problem, we propose a simple yet effective discriminative frequency domain-based FFN (DFFN), where we introduce a gated mechanism in the FFN based on the Joint Photographic Experts Group (JPEG) compression algorithm to discriminatively determine which low- and high-frequency information of the features should be preserved for latent clear image restoration. We formulate the proposed FSAS and DFFN into an asymmetrical network based on an encoder and decoder architecture, where the FSAS is only used in the decoder module for better image deblurring. Experimental results show that the proposed method performs favorably against the state-of-the-art approaches. Code will be available at \url{https://github.com/kkkls/FFTformer}.
The high-quality images yielded by generative adversarial networks (GANs) have motivated investigations into their application for image editing. However, GANs are often limited in the control they provide for performing specific edits. One of the principal challenges is the entangled latent space of GANs, which is not directly suitable for performing independent and detailed edits. Recent editing methods allow for either controlled style edits or controlled semantic edits. In addition, methods that use semantic masks to edit images have difficulty preserving the identity and are unable to perform controlled style edits. We propose a method to disentangle a GAN$\text{'}$s latent space into semantic and style spaces, enabling controlled semantic and style edits for face images independently within the same framework. To achieve this, we design an encoder-decoder based network architecture ($S^2$-Flow), which incorporates two proposed inductive biases. We show the suitability of $S^2$-Flow quantitatively and qualitatively by performing various semantic and style edits.
Wide-angle video sequences obtained by fisheye cameras exhibit characteristics that may not very well comply with standard image and video processing techniques such as error concealment. This paper introduces a temporal error concealment technique designed for the inherent characteristics of equisolid fisheye video sequences by applying a re-projection into the equisolid domain after conducting part of the error concealment in the perspective domain. Combining this technique with conventional decoder motion vector estimation achieves average gains of 0.71 dB compared against pure decoder motion vector estimation for the test sequences used. Maximum gains amount to up to 2.04 dB for selected frames.
Object detection is a computer vision task of predicting a set of bounding boxes and category labels for each object of interest in a given image. The category is related to a linguistic symbol such as 'dog' or 'person' and there should be relationships among them. However the object detector only learns to classify the categories and does not treat them as the linguistic symbols. Multi-modal models often use the pre-trained object detector to extract object features from the image, but the models are separated from the detector and the extracted visual features does not change with their linguistic input. We rethink the object detection as a vision-and-language reasoning task. We then propose targeted detection task, where detection targets are given by a natural language and the goal of the task is to detect only all the target objects in a given image. There are no detection if the target is not given. Commonly used modern object detectors have many hand-designed components like anchor and it is difficult to fuse the textual inputs into the complex pipeline. We thus propose Language-Targeted Detector (LTD) for the targeted detection based on a recently proposed Transformer-based detector. LTD is a encoder-decoder architecture and our conditional decoder allows the model to reason about the encoded image with the textual input as the linguistic context. We evaluate detection performances of LTD on COCO object detection dataset and also show that our model improves the detection results with the textual input grounding to the visual object.
Deep learning has been successful in BCI decoding. However, it is very data-hungry and requires pooling data from multiple sources. EEG data from various sources decrease the decoding performance due to negative transfer. Recently, transfer learning for EEG decoding has been suggested as a remedy and become subject to recent BCI competitions (e.g. BEETL), but there are two complications in combining data from many subjects. First, privacy is not protected as highly personal brain data needs to be shared (and copied across increasingly tight information governance boundaries). Moreover, BCI data are collected from different sources and are often based on different BCI tasks, which has been thought to limit their reusability. Here, we demonstrate a federated deep transfer learning technique, the Multi-dataset Federated Separate-Common-Separate Network (MF-SCSN) based on our previous work of SCSN, which integrates privacy-preserving properties into deep transfer learning to utilise data sets with different tasks. This framework trains a BCI decoder using different source data sets obtained from different imagery tasks (e.g. some data sets with hands and feet, vs others with single hands and tongue, etc). Therefore, by introducing privacy-preserving transfer learning techniques, we unlock the reusability and scalability of existing BCI data sets. We evaluated our federated transfer learning method on the NeurIPS 2021 BEETL competition BCI task. The proposed architecture outperformed the baseline decoder by 3%. Moreover, compared with the baseline and other transfer learning algorithms, our method protects the privacy of the brain data from different data centres.
Large-scale Transformer models bring significant improvements for various downstream vision language tasks with a unified architecture. The performance improvements come with increasing model size, resulting in slow inference speed and increased cost for severing. While some certain predictions benefit from the full complexity of the large-scale model, not all of inputs need the same amount of computation to conduct, potentially leading to computation resource waste. To handle this challenge, early exiting is proposed to adaptively allocate computational power in term of input complexity to improve inference efficiency. The existing early exiting strategies usually adopt output confidence based on intermediate layers as a proxy of input complexity to incur the decision of skipping following layers. However, such strategies cannot apply to encoder in the widely-used unified architecture with both encoder and decoder due to difficulty of output confidence estimation in the encoder. It is suboptimal in term of saving computation power to ignore the early exiting in encoder component. To handle this challenge, we propose a novel early exiting strategy for unified visual language models, which allows dynamically skip the layers in encoder and decoder simultaneously in term of input layer-wise similarities with multiple times of early exiting, namely \textbf{MuE}. By decomposing the image and text modalities in the encoder, MuE is flexible and can skip different layers in term of modalities, advancing the inference efficiency while minimizing performance drop. Experiments on the SNLI-VE and MS COCO datasets show that the proposed approach MuE can reduce expected inference time by up to 50\% and 40\% while maintaining 99\% and 96\% performance respectively.
The processing and recognition of geoscience images have wide applications. Most of existing researches focus on understanding the high-quality geoscience images by assuming that all the images are clear. However, in many real-world cases, the geoscience images might contain occlusions during the image acquisition. This problem actually implies the image inpainting problem in computer vision and multimedia. To the best of our knowledge, all the existing image inpainting algorithms learn to repair the occluded regions for a better visualization quality, they are excellent for natural images but not good enough for geoscience images by ignoring the geoscience related tasks. This paper aims to repair the occluded regions for a better geoscience task performance with the advanced visualization quality simultaneously, without changing the current deployed deep learning based geoscience models. Because of the complex context of geoscience images, we propose a coarse-to-fine encoder-decoder network with coarse-to-fine adversarial context discriminators to reconstruct the occluded image regions. Due to the limited data of geoscience images, we use a MaskMix based data augmentation method to exploit more information from limited geoscience image data. The experimental results on three public geoscience datasets for remote sensing scene recognition, cross-view geolocation and semantic segmentation tasks respectively show the effectiveness and accuracy of the proposed method.
End-to-end text spotting aims to integrate scene text detection and recognition into a unified framework. Dealing with the relationship between the two sub-tasks plays a pivotal role in designing effective spotters. Although Transformer-based methods eliminate the heuristic post-processing, they still suffer from the synergy issue between the sub-tasks and low training efficiency. In this paper, we present DeepSolo, a simple DETR-like baseline that lets a single Decoder with Explicit Points Solo for text detection and recognition simultaneously. Technically, for each text instance, we represent the character sequence as ordered points and model them with learnable explicit point queries. After passing a single decoder, the point queries have encoded requisite text semantics and locations, thus can be further decoded to the center line, boundary, script, and confidence of text via very simple prediction heads in parallel. Besides, we also introduce a text-matching criterion to deliver more accurate supervisory signals, thus enabling more efficient training. Quantitative experiments on public benchmarks demonstrate that DeepSolo outperforms previous state-of-the-art methods and achieves better training efficiency. In addition, DeepSolo is also compatible with line annotations, which require much less annotation cost than polygons. The code is available at https://github.com/ViTAE-Transformer/DeepSolo.
Camera relocalization has various applications in autonomous driving. Previous camera pose regression models consider only ideal scenarios where there is little environmental perturbation. To deal with challenging driving environments that may have changing seasons, weather, illumination, and the presence of unstable objects, we propose RobustLoc, which derives its robustness against perturbations from neural differential equations. Our model uses a convolutional neural network to extract feature maps from multi-view images, a robust neural differential equation diffusion block module to diffuse information interactively, and a branched pose decoder with multi-layer training to estimate the vehicle poses. Experiments demonstrate that RobustLoc surpasses current state-of-the-art camera pose regression models and achieves robust performance in various environments. Our code is released at: https://github.com/sijieaaa/RobustLoc
This paper presents an end-to-end neural mapping method for camera localization, dubbed NeuMap, encoding a whole scene into a grid of latent codes, with which a Transformer-based auto-decoder regresses 3D coordinates of query pixels. State-of-the-art feature matching methods require each scene to be stored as a 3D point cloud with per-point features, consuming several gigabytes of storage per scene. While compression is possible, performance drops significantly at high compression rates. Conversely, coordinate regression methods achieve high compression by storing scene information in a neural network but suffer from reduced robustness. NeuMap combines the advantages of both approaches by utilizing 1) learnable latent codes for efficient scene representation and 2) a scene-agnostic Transformer-based auto-decoder to infer coordinates for query pixels. This scene-agnostic network design learns robust matching priors from large-scale data and enables rapid optimization of codes for new scenes while keeping the network weights fixed. Extensive evaluations on five benchmarks show that NeuMap significantly outperforms other coordinate regression methods and achieves comparable performance to feature matching methods while requiring a much smaller scene representation size. For example, NeuMap achieves 39.1% accuracy in the Aachen night benchmark with only 6MB of data, whereas alternative methods require 100MB or several gigabytes and fail completely under high compression settings. The codes are available at https://github.com/Tangshitao/NeuMap
Federated learning (FL) is getting increased attention for processing sensitive, distributed datasets common to domains such as healthcare. Instead of directly training classification models on these datasets, recent works have considered training data generators capable of synthesising a new dataset which is not protected by any privacy restrictions. Thus, the synthetic data can be made available to anyone, which enables further evaluation of machine learning architectures and research questions off-site. As an additional layer of privacy-preservation, differential privacy can be introduced into the training process. We propose DPD-fVAE, a federated Variational Autoencoder with Differentially-Private Decoder, to synthesise a new, labelled dataset for subsequent machine learning tasks. By synchronising only the decoder component with FL, we can reduce the privacy cost per epoch and thus enable better data generators. In our evaluation on MNIST, Fashion-MNIST and CelebA, we show the benefits of DPD-fVAE and report competitive performance to related work in terms of Fr\'echet Inception Distance and accuracy of classifiers trained on the synthesised dataset.
Image cropping has progressed tremendously under the data-driven paradigm. However, current approaches do not account for the intentions of the user, which is an issue especially when the composition of the input image is complex. Moreover, labeling of cropping data is costly and hence the amount of data is limited, leading to poor generalization performance of current algorithms in the wild. In this work, we take advantage of vision-language models as a foundation for creating robust and user-intentional cropping algorithms. By adapting a transformer decoder with a pre-trained CLIP-based detection model, OWL-ViT, we develop a method to perform cropping with a text or image query that reflects the user's intention as guidance. In addition, our pipeline design allows the model to learn text-conditioned aesthetic cropping with a small cropping dataset, while inheriting the open-vocabulary ability acquired from millions of text-image pairs. We validate our model through extensive experiments on existing datasets as well as a new cropping test set we compiled that is characterized by content ambiguity.
Visual anomaly detection, an important problem in computer vision, is usually formulated as a one-class classification and segmentation task. The student-teacher (S-T) framework has proved to be effective in solving this challenge. However, previous works based on S-T only empirically applied constraints on normal data and fused multi-level information. In this study, we propose an improved model called DeSTSeg, which integrates a pre-trained teacher network, a denoising student encoder-decoder, and a segmentation network into one framework. First, to strengthen the constraints on anomalous data, we introduce a denoising procedure that allows the student network to learn more robust representations. From synthetically corrupted normal images, we train the student network to match the teacher network feature of the same images without corruption. Second, to fuse the multi-level S-T features adaptively, we train a segmentation network with rich supervision from synthetic anomaly masks, achieving a substantial performance improvement. Experiments on the industrial inspection benchmark dataset demonstrate that our method achieves state-of-the-art performance, 98.6% on image-level AUC, 75.8% on pixel-level average precision, and 76.4% on instance-level average precision.
Self-Supervised Video Hashing (SSVH) models learn to generate short binary representations for videos without ground-truth supervision, facilitating large-scale video retrieval efficiency and attracting increasing research attention. The success of SSVH lies in the understanding of video content and the ability to capture the semantic relation among unlabeled videos. Typically, state-of-the-art SSVH methods consider these two points in a two-stage training pipeline, where they firstly train an auxiliary network by instance-wise mask-and-predict tasks and secondly train a hashing model to preserve the pseudo-neighborhood structure transferred from the auxiliary network. This consecutive training strategy is inflexible and also unnecessary. In this paper, we propose a simple yet effective one-stage SSVH method called ConMH, which incorporates video semantic information and video similarity relationship understanding in a single stage. To capture video semantic information for better hashing learning, we adopt an encoder-decoder structure to reconstruct the video from its temporal-masked frames. Particularly, we find that a higher masking ratio helps video understanding. Besides, we fully exploit the similarity relationship between videos by maximizing agreement between two augmented views of a video, which contributes to more discriminative and robust hash codes. Extensive experiments on three large-scale video datasets (i.e., FCVID, ActivityNet and YFCC) indicate that ConMH achieves state-of-the-art results. Code is available at https://github.com/huangmozhi9527/ConMH.
Compression and reconstruction of visual data have been widely studied in the computer vision community, even before the popularization of deep learning. More recently, some have used deep learning to improve or refine existing pipelines, while others have proposed end-to-end approaches, including autoencoders and implicit neural representations, such as SIREN and NeRV. In this work, we propose Neural Visual Representation with Content-adaptive Embedding (CNeRV), which combines the generalizability of autoencoders with the simplicity and compactness of implicit representation. We introduce a novel content-adaptive embedding that is unified, concise, and internally (within-video) generalizable, that compliments a powerful decoder with a single-layer encoder. We match the performance of NeRV, a state-of-the-art implicit neural representation, on the reconstruction task for frames seen during training while far surpassing for frames that are skipped during training (unseen images). To achieve similar reconstruction quality on unseen images, NeRV needs 120x more time to overfit per-frame due to its lack of internal generalization. With the same latent code length and similar model size, CNeRV outperforms autoencoders on reconstruction of both seen and unseen images. We also show promising results for visual data compression. More details can be found in the project pagehttps://haochen-rye.github.io/CNeRV/
Aiming at the problem that the current video anomaly detection cannot fully use the temporal information and ignore the diversity of normal behavior, an anomaly detection method is proposed to integrate the spatiotemporal information of pedestrians. Based on the convolutional autoencoder, the input frame is compressed and restored through the encoder and decoder. Anomaly detection is realized according to the difference between the output frame and the true value. In order to strengthen the characteristic information connection between continuous video frames, the residual temporal shift module and the residual channel attention module are introduced to improve the modeling ability of the network on temporal information and channel information, respectively. Due to the excessive generalization of convolutional neural networks, in the memory enhancement modules, the hopping connections of each codec layer are added to limit autoencoders' ability to represent abnormal frames too vigorously and improve the anomaly detection accuracy of the network. In addition, the objective function is modified by a feature discretization loss, which effectively distinguishes different normal behavior patterns. The experimental results on the CUHK Avenue and ShanghaiTech datasets show that the proposed method is superior to the current mainstream video anomaly detection methods while meeting the real-time requirements.
In this report, we describe our submission to the Ego4D AudioVisual (AV) Speech Transcription Challenge 2022. Our pipeline is based on AVATAR, a state of the art encoder-decoder model for AV-ASR that performs early fusion of spectrograms and RGB images. We describe the datasets, experimental settings and ablations. Our final method achieves a WER of 68.40 on the challenge test set, outperforming the baseline by 43.7%, and winning the challenge.
We propose a Spiking Neural Network (SNN)-based explicit numerical scheme for long time integration of time-dependent Ordinary and Partial Differential Equations (ODEs, PDEs). The core element of the method is a SNN, trained to use spike-encoded information about the solution at previous timesteps to predict spike-encoded information at the next timestep. After the network has been trained, it operates as an explicit numerical scheme that can be used to compute the solution at future timesteps, given a spike-encoded initial condition. A decoder is used to transform the evolved spiking-encoded solution back to function values. We present results from numerical experiments of using the proposed method for ODEs and PDEs of varying complexity.
Generative models have recently started to outperform extractive models in Open Domain Question Answering, largely by leveraging their decoder to attend over multiple encoded passages and combining their information. However, generative models tend to be larger than extractive models due to the need for a decoder, run slower during inference due to auto-regressive decoder beam search, and their generated output often suffers from hallucinations. We propose to extend transformer encoders with the ability to fuse information from multiple passages, using global representation to provide cross-sample attention over all tokens across samples. Furthermore, we propose an alternative answer span probability calculation to better aggregate answer scores in the global space of all samples. Using our proposed method, we outperform the current state-of-the-art method by $2.5$ Exact Match score on the Natural Question dataset while using only $25\%$ of parameters and $35\%$ of the latency during inference, and $4.4$ Exact Match on WebQuestions dataset. When coupled with synthetic data augmentation, we outperform larger models on the TriviaQA dataset as well. The latency and parameter savings of our method make it particularly attractive for open-domain question answering, as these models are often compute-intensive.
Transformer, a model comprising attention-based encoder-decoder architecture, have gained prevalence in the field of natural language processing (NLP) and recently influenced the computer vision (CV) space. The similarities between computer vision and medical imaging, reviewed the question among researchers if the impact of transformers on computer vision be translated to medical imaging? In this paper, we attempt to provide a comprehensive and recent review on the application of transformers in medical imaging by; describing the transformer model comparing it with a diversity of convolutional neural networks (CNNs), detailing the transformer based approaches for medical image classification, segmentation, registration and reconstruction with a focus on the image modality, comparing the performance of state-of-the-art transformer architectures to best performing CNNs on standard medical datasets.