Models, code, and papers for "Jiahao Pang":
Inverse imaging problems are inherently under-determined, and hence it is important to employ appropriate image priors for regularization. One recent popular prior---the graph Laplacian regularizer---assumes that the target pixel patch is smooth with respect to an appropriately chosen graph. However, the mechanisms and implications of imposing the graph Laplacian regularizer on the original inverse problem are not well understood. To address this problem, in this paper we interpret neighborhood graphs of pixel patches as discrete counterparts of Riemannian manifolds and perform analysis in the continuous domain, providing insights into several fundamental aspects of graph Laplacian regularization for image denoising. Specifically, we first show the convergence of the graph Laplacian regularizer to a continuous-domain functional, integrating a norm measured in a locally adaptive metric space. Focusing on image denoising, we derive an optimal metric space assuming non-local self-similarity of pixel patches, leading to an optimal graph Laplacian regularizer for denoising in the discrete domain. We then interpret graph Laplacian regularization as an anisotropic diffusion scheme to explain its behavior during iterations, e.g., its tendency to promote piecewise smooth signals under certain settings. To verify our analysis, an iterative image denoising algorithm is developed. Experimental results show that our algorithm performs competitively with state-of-the-art denoising methods such as BM3D for natural images, and outperforms them significantly for piecewise smooth images.
Recently, it is increasingly popular to equip mobile RGB cameras with Time-of-Flight (ToF) sensors for active depth sensing. However, for off-the-shelf ToF sensors, one must tackle two problems in order to obtain high-quality depth with respect to the RGB camera, namely 1) online calibration and alignment; and 2) complicated error correction for ToF depth sensing. In this work, we propose a framework for jointly alignment and refinement via deep learning. First, a cross-modal optical flow between the RGB image and the ToF amplitude image is estimated for alignment. The aligned depth is then refined via an improved kernel predicting network that performs kernel normalization and applies the bias prior to the dynamic convolution. To enrich our data for end-to-end training, we have also synthesized a dataset using tools from computer graphics. Experimental results demonstrate the effectiveness of our approach, achieving state-of-the-art for ToF refinement.
With the developments of dual-lens camera modules,depth information representing the third dimension of thecaptured scenes becomes available for smartphones. It isestimated by stereo matching algorithms, taking as input thetwo views captured by dual-lens cameras at slightly differ-ent viewpoints. Depth-of-field rendering (also be referred toas synthetic defocus or bokeh) is one of the trending depth-based applications. However, to achieve fast depth estima-tion on smartphones, the stereo pairs need to be rectified inthe first place. In this paper, we propose a cost-effective so-lution to perform stereo rectification for dual-lens camerascalled direct self-rectification, short for DSR1. It removesthe need of individual offline calibration for every pair ofdual-lens cameras. In addition, the proposed solution isrobust to the slight movements, e.g., due to collisions, ofthe dual-lens cameras after fabrication. Different with ex-isting self-rectification approaches, our approach computesthe homography in a novel way with zero geometric distor-tions introduced to the master image. It is achieved by di-rectly minimizing the vertical displacements of correspond-ing points between the original master image and the trans-formed slave image. Our method is evaluated on both real-istic and synthetic stereo image pairs, and produces supe-rior results compared to the calibrated rectification or otherself-rectification approaches
We propose to combine the robustness merit of model-based approaches and the learning power of data-driven approaches for image restoration. Specifically, by integrating graph Laplacian regularization as a trainable module into a deep learning framework, we are less susceptible to overfitting than pure CNN-based approaches, achieving higher robustness to small dataset and cross-domain denoising. First, a sparse neighborhood graph is built from the output of a convolutional neural network (CNN). Then the image is restored by solving an unconstrained quadratic programming problem, using a corresponding graph Laplacian regularizer as a prior term. The proposed restoration pipeline is fully differentiable and hence can be end-to-end trained. Experimental results demonstrate that our work avoids overfitting given small training data. It is also endowed with strong cross-domain generalization power, outperforming the state-of-the-art approaches by remarkable margin.
3D point cloud - a new signal representation of volumetric objects - is a discrete collection of triples marking exterior object surface locations in 3D space. Conventional imperfect acquisition processes of 3D point cloud - e.g., stereo-matching from multiple viewpoint images or depth data acquired directly from active light sensors - imply non-negligible noise in the data. In this paper, we adopt a previously proposed low-dimensional manifold model for the surface patches in the point cloud and seek self-similar patches to denoise them simultaneously using the patch manifold prior. Due to discrete observations of the patches on the manifold, we approximate the manifold dimension computation defined in the continuous domain with a patch-based graph Laplacian regularizer and propose a new discrete patch distance measure to quantify the similarity between two same-sized surface patches for graph construction that is robust to noise. We show that our graph Laplacian regularizer has a natural graph spectral interpretation, and has desirable numerical stability properties via eigenanalysis. Extensive simulation results show that our proposed denoising scheme can outperform state-of-the-art methods in objective metrics and can better preserve visually salient structural features like edges.
Leveraging on the recent developments in convolutional neural networks (CNNs), matching dense correspondence from a stereo pair has been cast as a learning problem, with performance exceeding traditional approaches. However, it remains challenging to generate high-quality disparities for the inherently ill-posed regions. To tackle this problem, we propose a novel cascade CNN architecture composing of two stages. The first stage advances the recently proposed DispNet by equipping it with extra up-convolution modules, leading to disparity images with more details. The second stage explicitly rectifies the disparity initialized by the first stage; it couples with the first-stage and generates residual signals across multiple scales. The summation of the outputs from the two stages gives the final disparity. As opposed to directly learning the disparity at the second stage, we show that residual learning provides more effective refinement. Moreover, it also benefits the training of the overall cascade network. Experimentation shows that our cascade residual learning scheme provides state-of-the-art performance for matching stereo correspondence. By the time of the submission of this paper, our method ranks first in the KITTI 2015 stereo benchmark, surpassing the prior works by a noteworthy margin.
Previous monocular depth estimation methods take a single view and directly regress the expected results. Though recent advances are made by applying geometrically inspired loss functions during training, the inference procedure does not explicitly impose any geometrical constraint. Therefore these models purely rely on the quality of data and the effectiveness of learning to generalize. This either leads to suboptimal results or the demand of huge amount of expensive ground truth labelled data to generate reasonable results. In this paper, we show for the first time that the monocular depth estimation problem can be reformulated as two sub-problems, a view synthesis procedure followed by stereo matching, with two intriguing properties, namely i) geometrical constraints can be explicitly imposed during inference; ii) demand on labelled depth data can be greatly alleviated. We show that the whole pipeline can still be trained in an end-to-end fashion and this new formulation plays a critical role in advancing the performance. The resulting model outperforms all the previous monocular depth estimation methods as well as the stereo block matching method in the challenging KITTI dataset by only using a small number of real training data. The model also generalizes well to other monocular depth estimation benchmarks. We also discuss the implications and the advantages of solving monocular depth estimation using stereo methods.
Despite the recent success of stereo matching with convolutional neural networks (CNNs), it remains arduous to generalize a pre-trained deep stereo model to a novel domain. A major difficulty is to collect accurate ground-truth disparities for stereo pairs in the target domain. In this work, we propose a self-adaptation approach for CNN training, utilizing both synthetic training data (with ground-truth disparities) and stereo pairs in the new domain (without ground-truths). Our method is driven by two empirical observations. By feeding real stereo pairs of different domains to stereo models pre-trained with synthetic data, we see that: i) a pre-trained model does not generalize well to the new domain, producing artifacts at boundaries and ill-posed regions; however, ii) feeding an up-sampled stereo pair leads to a disparity map with extra details. To avoid i) while exploiting ii), we formulate an iterative optimization problem with graph Laplacian regularization. At each iteration, the CNN adapts itself better to the new domain: we let the CNN learn its own higher-resolution output; at the meanwhile, a graph Laplacian regularization is imposed to discriminatively keep the desired edges while smoothing out the artifacts. We demonstrate the effectiveness of our method in two domains: daily scenes collected by smartphone cameras, and street views captured in a driving car.
We observed that recent state-of-the-art results on single image human pose estimation were achieved by multi-stage Convolution Neural Networks (CNN). Notwithstanding the superior performance on static images, the application of these models on videos is not only computationally intensive, it also suffers from performance degeneration and flicking. Such suboptimal results are mainly attributed to the inability of imposing sequential geometric consistency, handling severe image quality degradation (e.g. motion blur and occlusion) as well as the inability of capturing the temporal correlation among video frames. In this paper, we proposed a novel recurrent network to tackle these problems. We showed that if we were to impose the weight sharing scheme to the multi-stage CNN, it could be re-written as a Recurrent Neural Network (RNN). This property decouples the relationship among multiple network stages and results in significantly faster speed in invoking the network for videos. It also enables the adoption of Long Short-Term Memory (LSTM) units between video frames. We found such memory augmented RNN is very effective in imposing geometric consistency among frames. It also well handles input quality degradation in videos while successfully stabilizes the sequential outputs. The experiments showed that our approach significantly outperformed current state-of-the-art methods on two large-scale video pose estimation benchmarks. We also explored the memory cells inside the LSTM and provided insights on why such mechanism would benefit the prediction for video-based pose estimations.
Recent advances in visual tracking showed that deep Convolutional Neural Networks (CNN) trained for image classification can be strong feature extractors for discriminative trackers. However, due to the drastic difference between image classification and tracking, extra treatments such as model ensemble and feature engineering must be carried out to bridge the two domains. Such procedures are either time consuming or hard to generalize well across datasets. In this paper we discovered that the internal structure of Region Proposal Network (RPN)'s top layer feature can be utilized for robust visual tracking. We showed that such property has to be unleashed by a novel loss function which simultaneously considers classification accuracy and bounding box quality. Without ensemble and any extra treatment on feature maps, our proposed method achieved state-of-the-art results on several large scale benchmarks including OTB50, OTB100 and VOT2016. We will make our code publicly available.
Most of the recent successful methods in accurate object detection and localization used some variants of R-CNN style two stage Convolutional Neural Networks (CNN) where plausible regions were proposed in the first stage then followed by a second stage for decision refinement. Despite the simplicity of training and the efficiency in deployment, the single stage detection methods have not been as competitive when evaluated in benchmarks consider mAP for high IoU thresholds. In this paper, we proposed a novel single stage end-to-end trainable object detection network to overcome this limitation. We achieved this by introducing Recurrent Rolling Convolution (RRC) architecture over multi-scale feature maps to construct object classifiers and bounding box regressors which are "deep in context". We evaluated our method in the challenging KITTI dataset which measures methods under IoU threshold of 0.7. We showed that with RRC, a single reduced VGG-16 based model already significantly outperformed all the previously published results. At the time this paper was written our models ranked the first in KITTI car detection (the hard level), the first in cyclist detection and the second in pedestrian detection. These results were not reached by the previous single stage methods. The code is publicly available.