We propose replacing scene text in videos using deep style transfer and learned photometric transformations.Building on recent progress on still image text replacement,we present extensions that alter text while preserving the appearance and motion characteristics of the original video.Compared to the problem of still image text replacement,our method addresses additional challenges introduced by video, namely effects induced by changing lighting, motion blur, diverse variations in camera-object pose over time,and preservation of temporal consistency. We parse the problem into three steps. First, the text in all frames is normalized to a frontal pose using a spatio-temporal trans-former network. Second, the text is replaced in a single reference frame using a state-of-art still-image text replacement method. Finally, the new text is transferred from the reference to remaining frames using a novel learned image transformation network that captures lighting and blur effects in a temporally consistent manner. Results on synthetic and challenging real videos show realistic text trans-fer, competitive quantitative and qualitative performance,and superior inference speed relative to alternatives. We introduce new synthetic and real-world datasets with paired text objects. To the best of our knowledge this is the first attempt at deep video text replacement.
The loss function of an unsupervised multimodal image registration framework has two terms, i.e., a metric for similarity measure and regularization. In the deep learning era, researchers proposed many approaches to automatically learn the similarity metric, which has been shown effective in improving registration performance. However, for the regularization term, most existing multimodal registration approaches still use a hand-crafted formula to impose artificial properties on the estimated deformation field. In this work, we propose a unimodal cyclic regularization training pipeline, which learns task-specific prior knowledge from simpler unimodal registration, to constrain the deformation field of multimodal registration. In the experiment of abdominal CT-MR registration, the proposed method yields better results over conventional regularization methods, especially for severely deformed local regions.
Magnetic resonance images (MRIs) are widely used to quantify vestibular schwannoma and the cochlea. Recently, deep learning methods have shown state-of-the-art performance for segmenting these structures. However, training segmentation models may require manual labels in target domain, which is expensive and time-consuming. To overcome this problem, domain adaptation is an effective way to leverage information from source domain to obtain accurate segmentations without requiring manual labels in target domain. In this paper, we propose an unsupervised learning framework to segment the VS and cochlea. Our framework leverages information from contrast-enhanced T1-weighted (ceT1-w) MRIs and its labels, and produces segmentations for T2-weighted MRIs without any labels in the target domain. We first applied a generator to achieve image-to-image translation. Next, we ensemble outputs from an ensemble of different models to obtain final segmentations. To cope with MRIs from different sites/scanners, we applied various 'online' augmentations during training to better capture the geometric variability and the variability in image appearance and quality. Our method is easy to build and produces promising segmentations, with a mean Dice score of 0.7930 and 0.7432 for VS and cochlea respectively in the validation set.
Gaussian mixture models find their place as a powerful tool, mostly in the clustering problem, but with proper preparation also in feature extraction, pattern recognition, image segmentation and in general machine learning. When faced with the problem of schema matching, different mixture models computed on different pieces of data can maintain crucial information about the structure of the dataset. In order to measure or compare results from mixture models, the Wasserstein distance can be very useful, however it is not easy to calculate for mixture distributions. In this paper we derive one of possible approximations for the Wasserstein distance between Gaussian mixture models and reduce it to linear problem. Furthermore, application examples concerning real world data are shown.
In medical image analysis, semi-supervised learning is an effective method to extract knowledge from a small amount of labeled data and a large amount of unlabeled data. This paper focuses on a popular pipeline known as self learning, and points out a weakness named lazy learning that refers to the difficulty for a model to learn from the pseudo labels generated by itself. To alleviate this issue, we propose ATSO, an asynchronous version of teacher-student optimization. ATSO partitions the unlabeled data into two subsets and alternately uses one subset to fine-tune the model and updates the label on the other subset. We evaluate ATSO on two popular medical image segmentation datasets and show its superior performance in various semi-supervised settings. With slight modification, ATSO transfers well to natural image segmentation for autonomous driving data.
Computer vision has achieved great success using standardized image representations -- pixel arrays, and the corresponding deep learning operators -- convolutions. In this work, we challenge this paradigm: we instead (a) represent images as a set of visual tokens and (b) apply visual transformers to find relationships between visual semantic concepts. Given an input image, we dynamically extract a set of visual tokens from the image to obtain a compact representation for high-level semantics. We then use visual transformers to operate over the visual tokens to densely model relationships between them. We find that this paradigm of token-based image representation and processing drastically outperforms its convolutional counterparts on image classification and semantic segmentation. To demonstrate the power of this approach on ImageNet classification, we use ResNet as a convenient baseline and use visual transformers to replace the last stage of convolutions. This reduces the stage's MACs by up to 6.9x, while attaining up to 4.53 points higher top-1 accuracy. For semantic segmentation, we use a visual-transformer-based FPN (VT-FPN) module to replace a convolution-based FPN, saving 6.5x fewer MACs while achieving up to 0.35 points higher mIoU on LIP and COCO-stuff.
Histopathological tissue classification is a fundamental task in pathomics cancer research. Precisely differentiating different tissue types is a benefit for the downstream researches, like cancer diagnosis, prognosis and etc. Existing works mostly leverage the popular classification backbones in computer vision to achieve histopathological tissue classification. In this paper, we proposed a super lightweight plug-and-play module, named Pyramidal Deep-Broad Learning (PDBL), for any well-trained classification backbone to further improve the classification performance without a re-training burden. We mimic how pathologists observe pathology slides in different magnifications and construct an image pyramid for the input image in order to obtain the pyramidal contextual information. For each level in the pyramid, we extract the multi-scale deep-broad features by our proposed Deep-Broad block (DB-block). We equipped PDBL in three popular classification backbones, ShuffLeNetV2, EfficientNetb0, and ResNet50 to evaluate the effectiveness and efficiency of our proposed module on two datasets (Kather Multiclass Dataset and the LC25000 Dataset). Experimental results demonstrate the proposed PDBL can steadily improve the tissue-level classification performance for any CNN backbones, especially for the lightweight models when given a small among of training samples (less than 10%), which greatly saves the computational time and annotation efforts.
In this paper, we study Multiscale Vision Transformers (MViT) as a unified architecture for image and video classification, as well as object detection. We present an improved version of MViT that incorporates decomposed relative positional embeddings and residual pooling connections. We instantiate this architecture in five sizes and evaluate it for ImageNet classification, COCO detection and Kinetics video recognition where it outperforms prior work. We further compare MViTs' pooling attention to window attention mechanisms where it outperforms the latter in accuracy/compute. Without bells-and-whistles, MViT has state-of-the-art performance in 3 domains: 88.8% accuracy on ImageNet classification, 56.1 box AP on COCO object detection as well as 86.1% on Kinetics-400 video classification. Code and models will be made publicly available.
The task of predicting smooth and edge-consistent depth maps is notoriously difficult for single image depth estimation. This paper proposes a novel Bilateral Grid based 3D convolutional neural network, dubbed as 3DBG-UNet, that parameterizes high dimensional feature space by encoding compact 3D bilateral grids with UNets and infers sharp geometric layout of the scene. Further, another novel 3DBGES-UNet model is introduced that integrate 3DBG-UNet for inferring an accurate depth map given a single color view. The 3DBGES-UNet concatenates 3DBG-UNet geometry map with the inception network edge accentuation map and a spatial object's boundary map obtained by leveraging semantic segmentation and train the UNet model with ResNet backbone. Both models are designed with a particular attention to explicitly account for edges or minute details. Preserving sharp discontinuities at depth edges is critical for many applications such as realistic integration of virtual objects in AR video or occlusion-aware view synthesis for 3D display applications.The proposed depth prediction network achieves state-of-the-art performance in both qualitative and quantitative evaluations on the challenging NYUv2-Depth data. The code and corresponding pre-trained weights will be made publicly available.
Convolutional Neural Networks, or CNNs, are undoubtedly the state of the art for image classification. However, they typically come with the cost of a large memory footprint. Recently, there has been significant progress in the field of image classification on memory-constrained devices, such as Arduino Unos, with novel contributions like the ProtoNN, Bonsai and FastGRNN models. These methods have been shown to perform excellently on tasks such as speech recognition or optical character recognition using MNIST, but their potential on more complex, multi-channel and multi-class image classification has yet to be determined. This paper presents a comprehensive analysis that shows that even in memory-constrained environments, CNNs implemented memory-optimally using Direct Convolutions outperform ProtoNN, Bonsai and FastGRNN models on 3-channel image classification using CIFAR-10. For our analysis, we propose new methods of adjusting the FastGRNN model to work with multi-channel images and then evaluate each algorithm with a memory size budget of 8KB, 16KB, 32KB, 64KB and 128KB to show quantitatively that CNNs are still state-of-the-art in image classification, even when memory size is constrained.