Diffusion models, praised for their success in generative tasks, are increasingly being applied to robotics, demonstrating exceptional performance in behavior cloning. However, their slow generation process stemming from iterative denoising steps poses a challenge for real-time applications in resource-constrained robotics setups and dynamically changing environments. In this paper, we introduce the One-Step Diffusion Policy (OneDP), a novel approach that distills knowledge from pre-trained diffusion policies into a single-step action generator, significantly accelerating response times for robotic control tasks. We ensure the distilled generator closely aligns with the original policy distribution by minimizing the Kullback-Leibler (KL) divergence along the diffusion chain, requiring only $2\%$-$10\%$ additional pre-training cost for convergence. We evaluated OneDP on 6 challenging simulation tasks as well as 4 self-designed real-world tasks using the Franka robot. The results demonstrate that OneDP not only achieves state-of-the-art success rates but also delivers an order-of-magnitude improvement in inference speed, boosting action prediction frequency from 1.5 Hz to 62 Hz, establishing its potential for dynamic and computationally constrained robotic applications. A video demo is provided at our project page, and the code will be publicly available.

The completion of large occluded human body images poses a unique challenge for general image completion methods. The complex shape variations of human bodies make it difficult to establish a consistent understanding of their structures. Furthermore, as human vision is highly sensitive to human bodies, even slight artifacts can significantly compromise image fidelity. To address these challenges, we propose a large occluded human image completion (LOHC) model based on a novel image-prior cooperative completion strategy. Our model leverages human segmentation maps as a prior, and completes the image and prior simultaneously. Compared to the widely adopted prior-then-image completion strategy for object completion, this cooperative completion process fosters more effective interaction between the prior and image information. Our model consists of two stages. The first stage is a transformer-based auto-regressive network that predicts the overall structure of the missing area by generating a coarse completed image at a lower resolution. The second stage is a convolutional network that refines the coarse images. As the coarse result may not always be accurate, we propose a Dynamic Fusion Module (DFM) to selectively fuses the useful features from the coarse image with the original input at spatial and channel levels. Through extensive experiments, we demonstrate our method’s superior performance compared to state-of-the-art methods.

Personalized text-to-image generation models enable users to create images that depict their individual possessions in diverse scenes, finding applications in various domains. To achieve the personalization capability, existing methods rely on finetuning a text-to-image foundation model on a user's custom dataset, which can be nontrivial for general users, resource-intensive, and time-consuming. Despite attempts to develope finetuning-free methods, their generation quality is much lower compared to their fine-tuning counterparts. In this paper, we propose Joint-Image Diffusion (JeDi), an effective technique for learning a finetuning-free personalization model. Our key idea is to learn the joint distribution of multiple related text-image pairs that share a common subject. To facilitate learning, we propose a scalable synthetic dataset generation technique. Once trained, our model enables fast and easy personalization at test time by simply using reference images as input during the sampling process. Our approach does not require any expensive optimization process or additional modules and can faithfully preserve the identity represented by any number of reference images. Experimental results show that our model achieves state-of-the-art gen-eration quality, both quantitatively and qualitatively, significantly outperforming both the prior finetuning-based and finetuning-free personalization baselines.

At the core of portrait photography is the search for ideal lighting and viewpoint. The process often requires advanced knowledge in photography and an elaborate studio setup. In this work, we propose Holo-Relighting, a volumetric relighting method that is capable of synthesizing novel viewpoints, and novel lighting from a single image. Holo-Relighting leverages the pretrained 3D GAN (EG3D) to reconstruct geometry and appearance from an input portrait as a set of 3D-aware features. We design a relighting module conditioned on a given lighting to process these features, and predict a relit 3D representation in the form of a tri-plane, which can render to an arbitrary viewpoint through volume rendering. Besides viewpoint and lighting control, Holo-Relighting also takes the head pose as a condition to enable head-pose-dependent lighting effects. With these novel designs, Holo-Relighting can generate complex non-Lambertian lighting effects (e.g., specular highlights and cast shadows) without using any explicit physical lighting priors. We train Holo-Relighting with data captured with a light stage, and propose two data-rendering techniques to improve the data quality for training the volumetric relighting system. Through quantitative and qualitative experiments, we demonstrate Holo-Relighting can achieve state-of-the-arts relighting quality with better photorealism, 3D consistency and controllability.

Existing image inpainting methods often produce artifacts that are caused by using vanilla convolution layers as building blocks that treat all image regions equally and generate holes at random locations with equal probability. This design does not differentiate the missing regions and valid regions in inference and does not consider the predictability of missing regions in training. To address these issues, we propose a deformable dynamic sampling (DDS) mechanism which is built on deformable convolutions (DCs), and a constraint is proposed to avoid the deformably sampled elements falling into the corrupted regions. Furthermore, to select both valid sample locations and suitable kernels dynamically, we equip DCs with content-aware dynamic kernel selection (DKS). In addition, to further encourage the DDS mechanism to find meaningful sampling locations, we propose to train the inpainting model with mined predictable regions as holes. During training, we jointly train a mask generator with the inpainting network to generate hole masks dynamically for each training sample. Thus, the mask generator can find large yet predictable missing regions as a better alternative to random masks. Extensive experiments demonstrate the advantages of our method over state-of-the-art methods qualitatively and quantitatively.

We propose a new framework for conditional image synthesis from semantic layouts of any precision levels, ranging from pure text to a 2D semantic canvas with precise shapes. More specifically, the input layout consists of one or more semantic regions with free-form text descriptions and adjustable precision levels, which can be set based on the desired controllability. The framework naturally reduces to text-to-image (T2I) at the lowest level with no shape information, and it becomes segmentation-to-image (S2I) at the highest level. By supporting the levels in-between, our framework is flexible in assisting users of different drawing expertise and at different stages of their creative workflow. We introduce several novel techniques to address the challenges coming with this new setup, including a pipeline for collecting training data; a precision-encoded mask pyramid and a text feature map representation to jointly encode precision level, semantics, and composition information; and a multi-scale guided diffusion model to synthesize images. To evaluate the proposed method, we collect a test dataset containing user-drawn layouts with diverse scenes and styles. Experimental results show that the proposed method can generate high-quality images following the layout at given precision, and compares favorably against existing methods. Project page https://zengxianyu.github.io/scenec/

Sketch-based image manipulation is an interactive image editing task to modify an image based on input sketches from users. Existing methods typically formulate this task as a conditional inpainting problem, which requires users to draw an extra mask indicating the region to modify in addition to sketches. The masked regions are regarded as holes and filled by an inpainting model conditioned on the sketch. With this formulation, paired training data can be easily obtained by randomly creating masks and extracting edges or contours. Although this setup simplifies data preparation and model design, it complicates user interaction and discards useful information in masked regions. To this end, we investigate a new paradigm of sketch-based image manipulation: mask-free local image manipulation, which only requires sketch inputs from users and utilizes the entire original image. Given an image and sketch, our model automatically predicts the target modification region and encodes it into a structure agnostic style vector. A generator then synthesizes the new image content based on the style vector and sketch. The manipulated image is finally produced by blending the generator output into the modification region of the original image. Our model can be trained in a self-supervised fashion by learning the reconstruction of an image region from the style vector and sketch. The proposed method offers simpler and more intuitive user workflows for sketch-based image manipulation and provides better results than previous approaches. More results, code and interactive demo will be available at https://zengxianyu.github.io/sketchedit.

High-cost pixel-level annotations makes it appealing to train saliency detection models with weak supervision. However, a single weak supervision source hardly contain enough information to train a well-performing model. To this end, we introduce a unified two-stage framework to learn from category labels, captions, web images and unlabeled images. In the first stage, we design a classification network (CNet) and a caption generation network (PNet), which learn to predict object categories and generate captions, respectively, meanwhile highlights the potential foreground regions. We present an attention transfer loss to transmit supervisions between two tasks and an attention coherence loss to encourage the networks to detect generally salient regions instead of task-specific regions. In the second stage, we create two complementary training datasets using CNet and PNet, i.e., natural image dataset with noisy labels for adapting saliency prediction network (SNet) to natural image input, and synthesized image dataset by pasting objects on background images for providing SNet with accurate ground-truth. During the testing phases, we only need SNet to predict saliency maps. Experiments indicate the performance of our method compares favorably against unsupervised, weakly supervised methods and even some supervised methods.

Recent deep generative inpainting methods use attention layers to allow the generator to explicitly borrow feature patches from the known region to complete a missing region. Due to the lack of supervision signals for the correspondence between missing regions and known regions, it may fail to find proper reference features, which often leads to artifacts in the results. Also, it computes pair-wise similarity across the entire feature map during inference bringing a significant computational overhead. To address this issue, we propose to teach such patch-borrowing behavior to an attention-free generator by joint training of an auxiliary contextual reconstruction task, which encourages the generated output to be plausible even when reconstructed by surrounding regions. The auxiliary branch can be seen as a learnable loss function, i.e. named as contextual reconstruction (CR) loss, where query-reference feature similarity and reference-based reconstructor are jointly optimized with the inpainting generator. The auxiliary branch ( i.e. CR loss) is required only during training, and only the inpainting generator is required during the inference. Experimental results demonstrate that the proposed inpainting model compares favourably against the state-of-the-art in terms of quantitative and visual performance. Code is available at https://github.com/zengxianyu/crfill.

Benefiting from the rapid development of Convolutional Neural Networks (CNNs), some salient object detection methods have achieved remarkable results by utilizing multi-level convolutional features. However, the saliency training datasets is of limited scale due to the high cost of pixel-level labeling, which leads to a limited generalization of the trained model on new scenarios during testing. Besides, some FCN-based methods directly integrate multi-level features, ignoring the fact that the noise in some features are harmful to saliency detection. In this paper, we propose a novel approach that transforms prior information into an embedding space to select attentive features and filter out outliers for salient object detection. Our network firstly generates a coarse prediction map through an encorder-decorder structure. Then a Feature Embedding Network (FEN) is trained to embed each pixel of the coarse map into a metric space, which incorporates much attentive features that highlight salient regions and suppress the response of non-salient regions. Further, the embedded features are refined through a deep-to-shallow Recursive Feature Integration Network (RFIN) to improve the details of prediction maps. Moreover, to alleviate the blurred boundaries, we propose a Guided Filter Refinement Network (GFRN) to jointly optimize the predicted results and the learnable guidance maps. Extensive experiments on five benchmark datasets demonstrate that our method outperforms state-of-the-art results. Our proposed method is end-to-end and achieves a realtime speed of 38 FPS.

The high cost of pixel-level annotations makes it appealing to train saliency detection models with weak supervision. However, a single weak supervision source usually does not contain enough information to train a well-performing model. To this end, we propose a unified framework to train saliency detection models with diverse weak supervision sources. In this paper, we use category labels, captions, and unlabelled data for training, yet other supervision sources can also be plugged into this flexible framework. We design a classification network (CNet) and a caption generation network (PNet), which learn to predict object categories and generate captions, respectively, meanwhile highlight the most important regions for corresponding tasks. An attention transfer loss is designed to transmit supervision signal between networks, such that the network designed to be trained with one supervision source can benefit from another. An attention coherence loss is defined on unlabelled data to encourage the networks to detect generally salient regions instead of task-specific regions. We use CNet and PNet to generate pixel-level pseudo labels to train a saliency prediction network (SNet). During the testing phases, we only need SNet to predict saliency maps. Experiments demonstrate the performance of our method compares favourably against unsupervised and weakly supervised methods and even some supervised methods.

Existing weakly supervised semantic segmentation (WSSS) methods usually utilize the results of pre-trained saliency detection (SD) models without explicitly modelling the connections between the two tasks, which is not the most efficient configuration. Here we propose a unified multi-task learning framework to jointly solve WSSS and SD using a single network, i.e. saliency and segmentation network (SSNet). SSNet consists of a segmentation network (SN) and a saliency aggregation module (SAM). For an input image, SN generates the segmentation result and, SAM predicts the saliency of each category and aggregating the segmentation masks of all categories into a saliency map. The proposed network is trained end-to-end with image-level category labels and class-agnostic pixel-level saliency labels. Experiments on PASCAL VOC 2012 segmentation dataset and four saliency benchmark datasets show the performance of our method compares favorably against state-of-the-art weakly supervised segmentation methods and fully supervised saliency detection methods.

The categories and appearance of salient objects vary from image to image, therefore, saliency detection is an image-specific task. Due to lack of large-scale saliency training data, using deep neural networks (DNNs) with pretraining is difficult to precisely capture the image-specific saliency cues. To solve this issue, we formulate a zero-shot learning problem to promote existing saliency detectors. Concretely, a DNN is trained as an embedding function to map pixels and the attributes of the salient/background regions of an image into the same metric space, in which an image-specific classifier is learned to classify the pixels. Since the image-specific task is performed by the classifier, the DNN embedding effectively plays the role of a general feature extractor. Compared with transferring the learning to a new recognition task using limited data, this formulation makes the DNN learn more effectively from small data. Extensive experiments on five data sets show that our method significantly improves accuracy of existing methods and compares favorably against state-of-the-art approaches.

We propose a novel unsupervised game-theoretic salient object detection algorithm that does not require labeled training data. First, saliency detection problem is formulated as a non-cooperative game, hereinafter referred to as Saliency Game, in which image regions are players who choose to be "background" or "foreground" as their pure strategies. A payoff function is constructed by exploiting multiple cues and combining complementary features. Saliency maps are generated according to each region's strategy in the Nash equilibrium of the proposed Saliency Game. Second, we explore the complementary relationship between color and deep features and propose an Iterative Random Walk algorithm to combine saliency maps produced by the Saliency Game using different features. Iterative random walk allows sharing information across feature spaces, and detecting objects that are otherwise very hard to detect. Extensive experiments over 6 challenging datasets demonstrate the superiority of our proposed unsupervised algorithm compared to several state of the art supervised algorithms.