Lazy Diffusion Transformer for Interactive Image Editing

We introduce a novel diffusion transformer, LazyDiffusion, that generates partial image updates efficiently. Our approach targets interactive image editing applications in which, starting from a blank canvas or an image, a user specifies a sequence of localized image modifications using binary masks and text prompts. Our generator operates in two phases. First, a context encoder processes the current canvas and user mask to produce a compact global context tailored to the region to generate. Second, conditioned on this context, a diffusion-based transformer decoder synthesizes the masked pixels in a "lazy" fashion, i.e., it only generates the masked region. This contrasts with previous works that either regenerate the full canvas, wasting time and computation, or confine processing to a tight rectangular crop around the mask, ignoring the global image context altogether. Our decoder's runtime scales with the mask size, which is typically small, while our encoder introduces negligible overhead. We demonstrate that our approach is competitive with state-of-the-art inpainting methods in terms of quality and fidelity while providing a 10x speedup for typical user interactions, where the editing mask represents 10% of the image.

Saliency-Aware Regularized Graph Neural Network

The crux of graph classification lies in the effective representation learning for the entire graph. Typical graph neural networks focus on modeling the local dependencies when aggregating features of neighboring nodes, and obtain the representation for the entire graph by aggregating node features. Such methods have two potential limitations: 1) the global node saliency w.r.t. graph classification is not explicitly modeled, which is crucial since different nodes may have different semantic relevance to graph classification; 2) the graph representation directly aggregated from node features may have limited effectiveness to reflect graph-level information. In this work, we propose the Saliency-Aware Regularized Graph Neural Network (SAR-GNN) for graph classification, which consists of two core modules: 1) a traditional graph neural network serving as the backbone for learning node features and 2) the Graph Neural Memory designed to distill a compact graph representation from node features of the backbone. We first estimate the global node saliency by measuring the semantic similarity between the compact graph representation and node features. Then the learned saliency distribution is leveraged to regularize the neighborhood aggregation of the backbone, which facilitates the message passing of features for salient nodes and suppresses the less relevant nodes. Thus, our model can learn more effective graph representation. We demonstrate the merits of SAR-GNN by extensive experiments on seven datasets across various types of graph data. Code will be released.

Edge-aware Plug-and-play Scheme for Semantic Segmentation

Semantic segmentation is a classic and fundamental computer vision problem dedicated to assigning each pixel with its corresponding class. Some recent methods introduce edge-based information for improving the segmentation performance. However these methods are specific and limited to certain network architectures, and they can not be transferred to other models or tasks. Therefore, we propose an abstract and universal edge supervision method called Edge-aware Plug-and-play Scheme (EPS), which can be easily and quickly applied to any semantic segmentation models. The core is edge-width/thickness preserving guided for semantic segmentation. The EPS first extracts the Edge Ground Truth (Edge GT) with a predefined edge thickness from the training data; and then for any network architecture, it directly copies the decoder head for the auxiliary task with the Edge GT supervision. To ensure the edge thickness preserving consistantly, we design a new boundarybased loss, called Polar Hausdorff (PH) Loss, for the auxiliary supervision. We verify the effectiveness of our EPS on the Cityscapes dataset using 22 models. The experimental results indicate that the proposed method can be seamlessly integrated into any state-of-the-art (SOTA) models with zero modification, resulting in promising enhancement of the segmentation performance.

Mask Focal Loss: A unifying framework for dense crowd counting with canonical object detection networks

As a fundamental computer vision task, crowd counting predicts the number of pedestrians in a scene, which plays an important role in risk perception and early warning, traffic control and scene statistical analysis. Currently, deep learning based head detection is a promising method for crowd counting. However, the highly concerned object detection networks cannot be well applied to this field for three reasons: (1) The sample imbalance has not been overcome yet in highly dense and complex scenes because the existing loss functions calculate the positive loss at a single key point or in the entire target area with the same weight for all pixels; (2) The canonical object detectors' loss calculation is a hard assignment without taking into account the space coherence from the object location to the background region; and (3) Most of the existing head detection datasets are only annotated with the center points instead of bounding boxes which is mandatory for the canonical detectors. To address these problems, we propose a novel loss function, called Mask Focal Loss (MFL), to redefine the loss contributions according to the situ value of the heatmap with a Gaussian kernel. MFL provides a unifying framework for the loss functions based on both heatmap and binary feature map ground truths. Meanwhile, for better evaluation and comparison, a new synthetic dataset GTA\_Head is built, including 35 sequences, 5096 images and 1732043 head labels with bounding boxes. Experimental results show the overwhelming performance and demonstrate that our proposed MFL framework is applicable to all of the canonical detectors and to various datasets with different annotation patterns. This work provides a strong baseline for surpassing the crowd counting methods based on density estimation.

Harmonizing Output Imbalance for semantic segmentation on extremely-imbalanced input data

Semantic segmentation is a high level computer vision task that assigns a label for each pixel of an image. It is challengeful to deal with extremely-imbalanced data in which the ratio of target ixels to background pixels is lower than 1:1000. Such severe input imbalance leads to output imbalance for poor model training. This paper considers three issues for extremely-imbalanced data: inspired by the region based loss, an implicit measure for the output imbalance is proposed, and an adaptive algorithm is designed for guiding the output imbalance hyperparameter selection; then it is generalized to distribution based loss for dealing with output imbalance; and finally a compound loss with our adaptive hyperparameter selection alogorithm can keep the consistency of training and inference for harmonizing the output imbalance. With four popular deep architectures on our private dataset with three input imbalance scales and three public datasets, extensive experiments demonstrate the ompetitive/promising performance of the proposed method.

Variational Distillation for Multi-View Learning

Information Bottleneck (IB) based multi-view learning provides an information theoretic principle for seeking shared information contained in heterogeneous data descriptions. However, its great success is generally attributed to estimate the multivariate mutual information which is intractable when the network becomes complicated. Moreover, the representation learning tradeoff, {\it i.e.}, prediction-compression and sufficiency-consistency tradeoff, makes the IB hard to satisfy both requirements simultaneously. In this paper, we design several variational information bottlenecks to exploit two key characteristics ({\it i.e.}, sufficiency and consistency) for multi-view representation learning. Specifically, we propose a Multi-View Variational Distillation (MV$^2$D) strategy to provide a scalable, flexible and analytical solution to fitting MI by giving arbitrary input of viewpoints but without explicitly estimating it. Under rigorously theoretical guarantee, our approach enables IB to grasp the intrinsic correlation between observations and semantic labels, producing predictive and compact representations naturally. Also, our information-theoretic constraint can effectively neutralize the sensitivity to heterogeneous data by eliminating both task-irrelevant and view-specific information, preventing both tradeoffs in multiple view cases. To verify our theoretically grounded strategies, we apply our approaches to various benchmarks under three different applications. Extensive experiments to quantitatively and qualitatively demonstrate the effectiveness of our approach against state-of-the-art methods.

Third Time's the Charm? Image and Video Editing with StyleGAN3

StyleGAN is arguably one of the most intriguing and well-studied generative models, demonstrating impressive performance in image generation, inversion, and manipulation. In this work, we explore the recent StyleGAN3 architecture, compare it to its predecessor, and investigate its unique advantages, as well as drawbacks. In particular, we demonstrate that while StyleGAN3 can be trained on unaligned data, one can still use aligned data for training, without hindering the ability to generate unaligned imagery. Next, our analysis of the disentanglement of the different latent spaces of StyleGAN3 indicates that the commonly used W/W+ spaces are more entangled than their StyleGAN2 counterparts, underscoring the benefits of using the StyleSpace for fine-grained editing. Considering image inversion, we observe that existing encoder-based techniques struggle when trained on unaligned data. We therefore propose an encoding scheme trained solely on aligned data, yet can still invert unaligned images. Finally, we introduce a novel video inversion and editing workflow that leverages the capabilities of a fine-tuned StyleGAN3 generator to reduce texture sticking and expand the field of view of the edited video.

StyleAlign: Analysis and Applications of Aligned StyleGAN Models

In this paper, we perform an in-depth study of the properties and applications of aligned generative models. We refer to two models as aligned if they share the same architecture, and one of them (the child) is obtained from the other (the parent) via fine-tuning to another domain, a common practice in transfer learning. Several works already utilize some basic properties of aligned StyleGAN models to perform image-to-image translation. Here, we perform the first detailed exploration of model alignment, also focusing on StyleGAN. First, we empirically analyze aligned models and provide answers to important questions regarding their nature. In particular, we find that the child model's latent spaces are semantically aligned with those of the parent, inheriting incredibly rich semantics, even for distant data domains such as human faces and churches. Second, equipped with this better understanding, we leverage aligned models to solve a diverse set of tasks. In addition to image translation, we demonstrate fully automatic cross-domain image morphing. We further show that zero-shot vision tasks may be performed in the child domain, while relying exclusively on supervision in the parent domain. We demonstrate qualitatively and quantitatively that our approach yields state-of-the-art results, while requiring only simple fine-tuning and inversion.

StyleCLIP: Text-Driven Manipulation of StyleGAN Imagery

Inspired by the ability of StyleGAN to generate highly realistic images in a variety of domains, much recent work has focused on understanding how to use the latent spaces of StyleGAN to manipulate generated and real images. However, discovering semantically meaningful latent manipulations typically involves painstaking human examination of the many degrees of freedom, or an annotated collection of images for each desired manipulation. In this work, we explore leveraging the power of recently introduced Contrastive Language-Image Pre-training (CLIP) models in order to develop a text-based interface for StyleGAN image manipulation that does not require such manual effort. We first introduce an optimization scheme that utilizes a CLIP-based loss to modify an input latent vector in response to a user-provided text prompt. Next, we describe a latent mapper that infers a text-guided latent manipulation step for a given input image, allowing faster and more stable text-based manipulation. Finally, we present a method for mapping a text prompts to input-agnostic directions in StyleGAN's style space, enabling interactive text-driven image manipulation. Extensive results and comparisons demonstrate the effectiveness of our approaches.

StyleSpace Analysis: Disentangled Controls for StyleGAN Image Generation

We explore and analyze the latent style space of StyleGAN2, a state-of-the-art architecture for image generation, using models pretrained on several different datasets. We first show that StyleSpace, the space of channel-wise style parameters, is significantly more disentangled than the other intermediate latent spaces explored by previous works. Next, we describe a method for discovering a large collection of style channels, each of which is shown to control a distinct visual attribute in a highly localized and disentangled manner. Third, we propose a simple method for identifying style channels that control a specific attribute, using a pretrained classifier or a small number of example images. Manipulation of visual attributes via these StyleSpace controls is shown to be better disentangled than via those proposed in previous works. To show this, we make use of a newly proposed Attribute Dependency metric. Finally, we demonstrate the applicability of StyleSpace controls to the manipulation of real images. Our findings pave the way to semantically meaningful and well-disentangled image manipulations via simple and intuitive interfaces.