An emerging family of language models (LMs), capable of processing both text and images within a single visual view, has the promise to unlock complex tasks such as chart understanding and UI navigation. We refer to these models as screenshot language models. Despite their appeal, existing screenshot LMs substantially lag behind text-only models on language understanding tasks. To close this gap, we adopt a simplified setting where the model inputs are plain-text-rendered screenshots, and we focus on improving the text ability of screenshot LMs. We propose a novel Patch-and-Text Prediction (PTP) objective, which masks and recovers both image patches of screenshots and text within screenshots. We also conduct extensive ablation studies on masking rates and patch sizes, as well as designs for improving training stability. Our pre-trained model, while solely taking visual inputs, achieves comparable performance with BERT on 6 out of 8 GLUE tasks (within 2%) and improves up to 8% over prior work. Additionally, we extend PTP to train autoregressive screenshot LMs and demonstrate its effectiveness--our models can significantly reduce perplexity by utilizing the screenshot context. Together, we hope our findings can inspire future research on developing powerful screenshot LMs and extending their reach to broader applications.
Generating high-quality videos that synthesize desired realistic content is a challenging task due to their intricate high-dimensionality and complexity of videos. Several recent diffusion-based methods have shown comparable performance by compressing videos to a lower-dimensional latent space, using traditional video autoencoder architecture. However, such method that employ standard frame-wise 2D and 3D convolution fail to fully exploit the spatio-temporal nature of videos. To address this issue, we propose a novel hybrid video diffusion model, called HVDM, which can capture spatio-temporal dependencies more effectively. The HVDM is trained by a hybrid video autoencoder which extracts a disentangled representation of the video including: (i) a global context information captured by a 2D projected latent (ii) a local volume information captured by 3D convolutions with wavelet decomposition (iii) a frequency information for improving the video reconstruction. Based on this disentangled representation, our hybrid autoencoder provide a more comprehensive video latent enriching the generated videos with fine structures and details. Experiments on video generation benchamarks (UCF101, SkyTimelapse, and TaiChi) demonstrate that the proposed approach achieves state-of-the-art video generation quality, showing a wide range of video applications (e.g., long video generation, image-to-video, and video dynamics control).
Chest X-Ray (CXR) is a widely used clinical imaging modality and has a pivotal role in the diagnosis and prognosis of various lung and heart related conditions. Conventional automated clinical diagnostic tool design strategies relying on radiology reads and supervised learning, entail the cumbersome requirement of high quality annotated training data. To address this challenge, self-supervised pre-training has proven to outperform supervised pre-training in numerous downstream vision tasks, representing a significant breakthrough in the field. However, medical imaging pre-training significantly differs from pre-training with natural images (e.g., ImageNet) due to unique attributes of clinical images. In this context, we introduce Diverse Concept Modeling (DiCoM), a novel self-supervised training paradigm that leverages a student teacher framework for learning diverse concepts and hence effective representation of the CXR data. Hence, expanding beyond merely modeling a single primary label within an image, instead, effectively harnessing the information from all the concepts inherent in the CXR. The pre-trained model is subsequently fine-tuned to address diverse domain-specific tasks. Our proposed paradigm consistently demonstrates robust performance across multiple downstream tasks on multiple datasets, highlighting the success and generalizability of the pre-training strategy. To establish the efficacy of our methods we analyze both the power of learned representations and the speed of convergence (SoC) of our models. For diverse data and tasks, DiCoM is able to achieve in most cases better results compared to other state-of-the-art pre-training strategies. This when combined with the higher SoC and generalization capabilities positions DiCoM to be established as a foundation model for CXRs, a widely used imaging modality.
This paper presents a neural architecture MVDiffusion++ for 3D object reconstruction that synthesizes dense and high-resolution views of an object given one or a few images without camera poses. MVDiffusion++ achieves superior flexibility and scalability with two surprisingly simple ideas: 1) A ``pose-free architecture'' where standard self-attention among 2D latent features learns 3D consistency across an arbitrary number of conditional and generation views without explicitly using camera pose information; and 2) A ``view dropout strategy'' that discards a substantial number of output views during training, which reduces the training-time memory footprint and enables dense and high-resolution view synthesis at test time. We use the Objaverse for training and the Google Scanned Objects for evaluation with standard novel view synthesis and 3D reconstruction metrics, where MVDiffusion++ significantly outperforms the current state of the arts. We also demonstrate a text-to-3D application example by combining MVDiffusion++ with a text-to-image generative model.
Image restoration tasks traditionally rely on convolutional neural networks. However, given the local nature of the convolutional operator, they struggle to capture global information. The promise of attention mechanisms in Transformers is to circumvent this problem, but it comes at the cost of intensive computational overhead. Many recent studies in image restoration have focused on solving the challenge of balancing performance and computational cost via Transformer variants. In this paper, we present CascadedGaze Network (CGNet), an encoder-decoder architecture that employs Global Context Extractor (GCE), a novel and efficient way to capture global information for image restoration. The GCE module leverages small kernels across convolutional layers to learn global dependencies, without requiring self-attention. Extensive experimental results show that our approach outperforms a range of state-of-the-art methods on denoising benchmark datasets including both real image denoising and synthetic image denoising, as well as on image deblurring task, while being more computationally efficient.
Augmented reality for laparoscopic liver resection is a visualisation mode that allows a surgeon to localise tumours and vessels embedded within the liver by projecting them on top of a laparoscopic image. Preoperative 3D models extracted from CT or MRI data are registered to the intraoperative laparoscopic images during this process. In terms of 3D-2D fusion, most of the algorithms make use of anatomical landmarks to guide registration. These landmarks include the liver's inferior ridge, the falciform ligament, and the occluding contours. They are usually marked by hand in both the laparoscopic image and the 3D model, which is time-consuming and may contain errors if done by a non-experienced user. Therefore, there is a need to automate this process so that augmented reality can be used effectively in the operating room. We present the Preoperative-to-Intraoperative Laparoscopic Fusion Challenge (P2ILF), held during the Medical Imaging and Computer Assisted Interventions (MICCAI 2022) conference, which investigates the possibilities of detecting these landmarks automatically and using them in registration. The challenge was divided into two tasks: 1) A 2D and 3D landmark detection task and 2) a 3D-2D registration task. The teams were provided with training data consisting of 167 laparoscopic images and 9 preoperative 3D models from 9 patients, with the corresponding 2D and 3D landmark annotations. A total of 6 teams from 4 countries participated, whose proposed methods were evaluated on 16 images and two preoperative 3D models from two patients. All the teams proposed deep learning-based methods for the 2D and 3D landmark segmentation tasks and differentiable rendering-based methods for the registration task. Based on the experimental outcomes, we propose three key hypotheses that determine current limitations and future directions for research in this domain.
Contemporary cutting-edge open-vocabulary segmentation approaches commonly rely on image-mask-text triplets, yet this restricted annotation is labour-intensive and encounters scalability hurdles in complex real-world scenarios. Although some methods are proposed to reduce the annotation cost with only text supervision, the incompleteness of supervision severely limits the versatility and performance. In this paper, we liberate the strict correspondence between masks and texts by using independent image-mask and image-text pairs, which can be easily collected respectively. With this unpaired mask-text supervision, we propose a new weakly-supervised open-vocabulary segmentation framework (Uni-OVSeg) that leverages confident pairs of mask predictions and entities in text descriptions. Using the independent image-mask and image-text pairs, we predict a set of binary masks and associate them with entities by resorting to the CLIP embedding space. However, the inherent noise in the correspondence between masks and entities poses a significant challenge when obtaining reliable pairs. In light of this, we advocate using the large vision-language model (LVLM) to refine text descriptions and devise a multi-scale ensemble to stablise the matching between masks and entities. Compared to text-only weakly-supervised methods, our Uni-OVSeg achieves substantial improvements of 15.5% mIoU on the ADE20K datasets, and even surpasses fully-supervised methods on the challenging PASCAL Context-459 dataset.
Digital watermarking enables protection against copyright infringement of images. Although existing methods embed watermarks imperceptibly and demonstrate robustness against attacks, they typically lack resilience against geometric transformations. Therefore, this paper proposes a new watermarking method that is robust against geometric attacks. The proposed method is based on the existing HiDDeN architecture that uses deep learning for watermark encoding and decoding. We add new noise layers to this architecture, namely for a differentiable JPEG estimation, rotation, rescaling, translation, shearing and mirroring. We demonstrate that our method outperforms the state of the art when it comes to geometric robustness. In conclusion, the proposed method can be used to protect images when viewed on consumers' devices.
Sampling from diffusion probabilistic models (DPMs) is often expensive for high-quality image generation and typically requires many steps with a large model. In this paper, we introduce sampling Trajectory Stitching T-Stitch, a simple yet efficient technique to improve the sampling efficiency with little or no generation degradation. Instead of solely using a large DPM for the entire sampling trajectory, T-Stitch first leverages a smaller DPM in the initial steps as a cheap drop-in replacement of the larger DPM and switches to the larger DPM at a later stage. Our key insight is that different diffusion models learn similar encodings under the same training data distribution and smaller models are capable of generating good global structures in the early steps. Extensive experiments demonstrate that T-Stitch is training-free, generally applicable for different architectures, and complements most existing fast sampling techniques with flexible speed and quality trade-offs. On DiT-XL, for example, 40% of the early timesteps can be safely replaced with a 10x faster DiT-S without performance drop on class-conditional ImageNet generation. We further show that our method can also be used as a drop-in technique to not only accelerate the popular pretrained stable diffusion (SD) models but also improve the prompt alignment of stylized SD models from the public model zoo. Code is released at https://github.com/NVlabs/T-Stitch
Animating virtual characters has always been a fundamental research problem in virtual reality (VR). Facial animations play a crucial role as they effectively convey emotions and attitudes of virtual humans. However, creating such facial animations can be challenging, as current methods often involve utilization of expensive motion capture devices or significant investments of time and effort from human animators in tuning animation parameters. In this paper, we propose a holistic solution to automatically animate virtual human faces. In our solution, a deep learning model was first trained to retarget the facial expression from input face images to virtual human faces by estimating the blendshape coefficients. This method offers the flexibility of generating animations with characters of different appearances and blendshape topologies. Second, a practical toolkit was developed using Unity 3D, making it compatible with the most popular VR applications. The toolkit accepts both image and video as input to animate the target virtual human faces and enables users to manipulate the animation results. Furthermore, inspired by the spirit of Human-in-the-loop (HITL), we leveraged user feedback to further improve the performance of the model and toolkit, thereby increasing the customization properties to suit user preferences. The whole solution, for which we will make the code public, has the potential to accelerate the generation of facial animations for use in VR applications.