Vision transformers have shown great success on numerous computer vision tasks. However, its central component, softmax attention, prohibits vision transformers from scaling up to high-resolution images, due to both the computational complexity and memory footprint being quadratic. Although linear attention was introduced in natural language processing (NLP) tasks to mitigate a similar issue, directly applying existing linear attention to vision transformers may not lead to satisfactory results. We investigate this problem and find that computer vision tasks focus more on local information compared with NLP tasks. Based on this observation, we present a Vicinity Attention that introduces a locality bias to vision transformers with linear complexity. Specifically, for each image patch, we adjust its attention weight based on its 2D Manhattan distance measured by its neighbouring patches. In this case, the neighbouring patches will receive stronger attention than far-away patches. Moreover, since our Vicinity Attention requires the token length to be much larger than the feature dimension to show its efficiency advantages, we further propose a new Vicinity Vision Transformer (VVT) structure to reduce the feature dimension without degenerating the accuracy. We perform extensive experiments on the CIFAR100, ImageNet1K, and ADE20K datasets to validate the effectiveness of our method. Our method has a slower growth rate of GFlops than previous transformer-based and convolution-based networks when the input resolution increases. In particular, our approach achieves state-of-the-art image classification accuracy with 50% fewer parameters than previous methods.
Technological and computational advances continuously drive forward the broad field of deep learning. In recent years, the derivation of quantities describing theuncertainty in the prediction - which naturally accompanies the modeling process - has sparked general interest in the deep learning community. Often neglected in the machine learning setting is the human uncertainty that influences numerous labeling processes. As the core of this work, label uncertainty is explicitly embedded into the training process via distributional labels. We demonstrate the effectiveness of our approach on image classification with a remote sensing data set that contains multiple label votes by domain experts for each image: The incorporation of label uncertainty helps the model to generalize better to unseen data and increases model performance. Similar to existing calibration methods, the distributional labels lead to better-calibrated probabilities, which in turn yield more certain and trustworthy predictions.
We present an algorithm for re-rendering a person from a single image under arbitrary poses. Existing methods often have difficulties in hallucinating occluded contents photo-realistically while preserving the identity and fine details in the source image. We first learn to inpaint the correspondence field between the body surface texture and the source image with a human body symmetry prior. The inpainted correspondence field allows us to transfer/warp local features extracted from the source to the target view even under large pose changes. Directly mapping the warped local features to an RGB image using a simple CNN decoder often leads to visible artifacts. Thus, we extend the StyleGAN generator so that it takes pose as input (for controlling poses) and introduces a spatially varying modulation for the latent space using the warped local features (for controlling appearances). We show that our method compares favorably against the state-of-the-art algorithms in both quantitative evaluation and visual comparison.
Quantitative magnetic resonance (MR) parametric mapping is a promising approach for characterizing intrinsic tissue-dependent information. However, long scan time significantly hinders its widespread applications. Recently, low-rank tensor has been employed and demonstrated good performance in accelerating MR parametricmapping. In this study, we propose a novel method that uses spatial patch-based and parametric group-based low rank tensors simultaneously (SMART) to reconstruct images from highly undersampled k-space data. The spatial patch-based low-rank tensor exploits the high local and nonlocal redundancies and similarities between the contrast images in parametric mapping. The parametric group based low-rank tensor, which integrates similar exponential behavior of the image signals, is jointly used to enforce the multidimensional low-rankness in the reconstruction process. In vivo brain datasets were used to demonstrate the validity of the proposed method. Experimental results have demonstrated that the proposed method achieves 11.7-fold and 13.21-fold accelerations in two-dimensional and three-dimensional acquisitions, respectively, with more accurate reconstructed images and maps than several state-of-the-art methods. Prospective reconstruction results further demonstrate the capability of the SMART method in accelerating MR quantitative imaging.
Regression plays an essential role in many medical imaging applications for estimating various clinical risk or measurement scores. While training strategies and loss functions have been studied for the deep neural networks in medical image classification tasks, options for regression tasks are very limited. One of the key challenges is that the high-dimensional feature representation learned by existing popular loss functions like Mean Squared Error or L1 loss is hard to interpret. In this paper, we propose a novel Regression Metric Loss (RM-Loss), which endows the representation space with the semantic meaning of the label space by finding a representation manifold that is isometric to the label space. Experiments on two regression tasks, i.e. coronary artery calcium score estimation and bone age assessment, show that RM-Loss is superior to the existing popular regression losses on both performance and interpretability. Code is available at https://github.com/DIAL-RPI/Regression-Metric-Loss.
Person reidentification (re-ID) has been receiving increasing attention in recent years due to its importance for both science and society. Machine learning and particularly Deep Learning (DL) has become the main re-id tool that allowed researches to achieve unprecedented accuracy levels on benchmark datasets. However, there is a known problem of poor generalization of DL models. That is, models trained to achieve high accuracy on one dataset perform poorly on other ones and require re-training. To address this issue, we present a model without trainable parameters which shows great potential for high generalization. It combines a fully analytical feature extraction and similarity ranking scheme with DL-based human parsing used to obtain the initial subregion classification. We show that such combination to a high extent eliminates the drawbacks of existing analytical methods. We use interpretable color and texture features which have human-readable similarity measures associated with them. To verify the proposed method we conduct experiments on Market1501 and CUHK03 datasets achieving competitive rank-1 accuracy comparable with that of DL-models. Most importantly we show that our method achieves 63.9% and 93.5% rank-1 cross-domain accuracy when applied to transfer learning tasks. It is significantly higher than previously reported 30-50% transfer accuracy. We discuss the potential ways of adding new features to further improve the model. We also show the advantage of interpretable features for constructing human-generated queries from verbal description to conduct search without a query image.
Recently, image-to-image translation has made significant progress in achieving both multi-label (\ie, translation conditioned on different labels) and multi-style (\ie, generation with diverse styles) tasks. However, due to the unexplored independence and exclusiveness in the labels, existing endeavors are defeated by involving uncontrolled manipulations to the translation results. In this paper, we propose Hierarchical Style Disentanglement (HiSD) to address this issue. Specifically, we organize the labels into a hierarchical tree structure, in which independent tags, exclusive attributes, and disentangled styles are allocated from top to bottom. Correspondingly, a new translation process is designed to adapt the above structure, in which the styles are identified for controllable translations. Both qualitative and quantitative results on the CelebA-HQ dataset verify the ability of the proposed HiSD. We hope our method will serve as a solid baseline and provide fresh insights with the hierarchically organized annotations for future research in image-to-image translation. The code has been released at https://github.com/imlixinyang/HiSD.
Trained using only image class label, deep weakly supervised methods allow image classification and ROI segmentation for interpretability. Despite their success on natural images, they face several challenges over histology data where ROI are visually similar to background making models vulnerable to high pixel-wise false positives. These methods lack mechanisms for modeling explicitly non-discriminative regions which raises false-positive rates. We propose novel regularization terms, which enable the model to seek both non-discriminative and discriminative regions, while discouraging unbalanced segmentations and using only image class label. Our method is composed of two networks: a localizer that yields segmentation mask, followed by a classifier. The training loss pushes the localizer to build a segmentation mask that holds most discrimiantive regions while simultaneously modeling background regions. Comprehensive experiments over two histology datasets showed the merits of our method in reducing false positives and accurately segmenting ROI.
Image classification is an important task in various machine learning applications. In recent years, a number of classification methods based on quantum machine learning and different quantum image encoding techniques have been proposed. In this paper, we study the effect of three different quantum image encoding approaches on the performance of a convolution-inspired hybrid quantum-classical image classification algorithm called quanvolutional neural network (QNN). We furthermore examine the effect of variational - i.e. trainable - quantum circuits on the classification results. Our experiments indicate that some image encodings are better suited for variational circuits. However, our experiments show as well that there is not one best image encoding, but that the choice of the encoding depends on the specific constraints of the application.
We present NeuriCam, a key-frame video super-resolution and colorization based system, to achieve low-power video capture from dual-mode IOT cameras. Our idea is to design a dual-mode camera system where the first mode is low power (1.1~mW) but only outputs gray-scale, low resolution and noisy video and the second mode consumes much higher power (100~mW) but outputs color and higher resolution images. To reduce total energy consumption, we heavily duty cycle the high power mode to output an image only once every second. The data from this camera system is then wirelessly streamed to a nearby plugged-in gateway, where we run our real-time neural network decoder to reconstruct a higher resolution color video. To achieve this, we introduce an attention feature filter mechanism that assigns different weights to different features, based on the correlation between the feature map and contents of the input frame at each spatial location. We design a wireless hardware prototype using off-the-shelf cameras and address practical issues including packet loss and perspective mismatch. Our evaluation shows that our dual-camera hardware reduces camera energy consumption while achieving an average gray-scale PSNR gain of 3.7~dB over prior video super resolution methods and 5.6~dB RGB gain over existing color propagation methods. Open-source code: https://github.com/vb000/NeuriCam.