Jeffrey Zhang

Kedan Li, Jeffrey Zhang, Shao-Yu Chang et al.

An outfit visualization method generates an image of a person wearing real garments from images of those garments. Current methods can produce images that look realistic and preserve garment identity, captured in details such as collar, cuffs, texture, hem, and sleeve length. However, no current method can both control how the garment is worn -- including tuck or untuck, opened or closed, high or low on the waist, etc.. -- and generate realistic images that accurately preserve the properties of the original garment. We describe an outfit visualization method that controls drape while preserving garment identity. Our system allows instance independent editing of garment drape, which means a user can construct an edit (e.g. tucking a shirt in a specific way) that can be applied to all shirts in a garment collection. Garment detail is preserved by relying on a warping procedure to place the garment on the body and a generator then supplies fine shading detail. To achieve instance independent control, we use control points with garment category-level semantics to guide the warp. The method produces state-of-the-art quality images, while allowing creative ways to style garments, including allowing tops to be tucked or untucked; jackets to be worn open or closed; skirts to be worn higher or lower on the waist; and so on. The method allows interactive control to correct errors in individual renderings too. Because the edits are instance independent, they can be applied to large pools of garments automatically and can be conditioned on garment metadata (e.g. all cropped jackets are worn closed or all bomber jackets are worn closed).

Qianqian Xie, Qingyu Chen, Aokun Chen et al.

Recent advancements in large language models (LLMs) like ChatGPT and LLaMA show promise in medical applications, yet challenges remain in medical language comprehension. This study presents Me-LLaMA, a new medical LLM family based on open-source LLaMA models, optimized for medical text analysis and diagnosis by leveraging large-scale, domain-specific datasets. The Me-LLaMA family, including foundation models Me-LLaMA 13/70B and their chat-enhanced versions, was developed through continued pre-training and instruction tuning with 129B tokens and 214K samples from biomedical and clinical sources. Training the 70B models required over 100,000 A100 GPU hours. Me-LLaMA's performance was evaluated across six medical text analysis tasks using 12 benchmark datasets and complex clinical case diagnosis, with automatic and human evaluations. Results indicate Me-LLaMA outperforms LLaMA and other open-source medical LLMs in zero-shot and supervised settings. Task-specific tuning further boosts performance, surpassing ChatGPT on 7 of 8 datasets and GPT-4 on 5 of 8. For complex clinical cases, Me-LLaMA achieves performance comparable to ChatGPT and GPT-4. This work underscores the importance of domain-specific data in developing medical LLMs and addresses the high computational costs involved in training, highlighting a balance between pre-training and fine-tuning strategies. Me-LLaMA models are now accessible under user agreements, providing a valuable resource for advancing medical AI.

Amir Ali Ahmadi, Abraar Chaudhry, Jeffrey Zhang

We present generalizations of Newton's method that incorporate derivatives of an arbitrary order $d$ but maintain a polynomial dependence on dimension in their cost per iteration. At each step, our $d^{\text{th}}$-order method uses semidefinite programming to construct and minimize a sum of squares-convex approximation to the $d^{\text{th}}$-order Taylor expansion of the function we wish to minimize. We prove that our $d^{\text{th}}$-order method has local convergence of order $d$. This results in lower oracle complexity compared to the classical Newton method. We show on numerical examples that basins of attraction around local minima can get larger as $d$ increases. Under additional assumptions, we present a modified algorithm, again with polynomial cost per iteration, which is globally convergent and has local convergence of order $d$.

Jeffrey Zhang, Shao-Yu Chang, Kedan Li et al.

Retail photography imposes specific requirements on images. For instance, images may need uniform background colors, consistent model poses, centered products, and consistent lighting. Minor deviations from these standards impact a site's aesthetic appeal, making the images unsuitable for use. We show that Stable Diffusion methods, as currently applied, do not respect these requirements. The usual practice of training the denoiser with a very noisy image and starting inference with a sample of pure noise leads to inconsistent generated images during inference. This inconsistency occurs because it is easy to tell the difference between samples of the training and inference distributions. As a result, a network trained with centered retail product images with uniform backgrounds generates images with erratic backgrounds. The problem is easily fixed by initializing inference with samples from an approximation of noisy images. However, in using such an approximation, the joint distribution of text and noisy image at inference time still slightly differs from that at training time. This discrepancy is corrected by training the network with samples from the approximate noisy image distribution. Extensive experiments on real application data show significant qualitative and quantitative improvements in performance from adopting these procedures. Finally, our procedure can interact well with other control-based methods to further enhance the controllability of diffusion-based methods.

Jeffrey Zhang, Kedan Li, Shao-Yu Chang et al.

Virtual Try-on (VTON) involves generating images of a person wearing selected garments. Diffusion-based methods, in particular, can create high-quality images, but they struggle to maintain the identities of the input garments. We identified this problem stems from the specifics in the training formulation for diffusion. To address this, we propose a unique training scheme that limits the scope in which diffusion is trained. We use a control image that perfectly aligns with the target image during training. In turn, this accurately preserves garment details during inference. We demonstrate our method not only effectively conserves garment details but also allows for layering, styling, and shoe try-on. Our method runs multi-garment try-on in a single inference cycle and can support high-quality zoomed-in generations without training in higher resolutions. Finally, we show our method surpasses prior methods in accuracy and quality.

Stefan Bauer, Felix Widmaier, Manuel Wüthrich et al.

Dexterous manipulation remains an open problem in robotics. To coordinate efforts of the research community towards tackling this problem, we propose a shared benchmark. We designed and built robotic platforms that are hosted at MPI for Intelligent Systems and can be accessed remotely. Each platform consists of three robotic fingers that are capable of dexterous object manipulation. Users are able to control the platforms remotely by submitting code that is executed automatically, akin to a computational cluster. Using this setup, i) we host robotics competitions, where teams from anywhere in the world access our platforms to tackle challenging tasks ii) we publish the datasets collected during these competitions (consisting of hundreds of robot hours), and iii) we give researchers access to these platforms for their own projects.

Kedan Li, Min jin Chong, Jeffrey Zhang et al.

Virtual try-on methods aim to generate images of fashion models wearing arbitrary combinations of garments. This is a challenging task because the generated image must appear realistic and accurately display the interaction between garments. Prior works produce images that are filled with artifacts and fail to capture important visual details necessary for commercial applications. We propose Outfit Visualization Net (OVNet) to capture these important details (e.g. buttons, shading, textures, realistic hemlines, and interactions between garments) and produce high quality multiple-garment virtual try-on images. OVNet consists of 1) a semantic layout generator and 2) an image generation pipeline using multiple coordinated warps. We train the warper to output multiple warps using a cascade loss, which refines each successive warp to focus on poorly generated regions of a previous warp and yields consistent improvements in detail. In addition, we introduce a method for matching outfits with the most suitable model and produce significant improvements for both our and other previous try-on methods. Through quantitative and qualitative analysis, we demonstrate our method generates substantially higher-quality studio images compared to prior works for multi-garment outfits. An interactive interface powered by this method has been deployed on fashion e-commerce websites and received overwhelmingly positive feedback.

Claire Chen, Krishnan Srinivasan, Jeffrey Zhang et al.

This report describes our approach for Phase 3 of the Real Robot Challenge. To solve cuboid manipulation tasks of varying difficulty, we decompose each task into the following primitives: moving the fingers to the cuboid to grasp it, turning it on the table to minimize orientation error, and re-positioning it to the goal position. We use model-based trajectory optimization and control to plan and execute these primitives. These grasping, turning, and re-positioning primitives are sequenced with a state-machine that determines which primitive to execute given the current object state and goal. Our method shows robust performance over multiple runs with randomized initial and goal positions. With this approach, our team placed second in the challenge, under the anonymous name "sombertortoise" on the leaderboard. Example runs of our method solving each of the four levels can be seen in this video (https://www.youtube.com/watch?v=I65Kwu9PGmg&list=PLt9QxrtaftrHGXcp4Oh8-s_OnQnBnLtei&index=1).

Jeffrey Zhang

In this thesis, we settle the computational complexity of some fundamental questions in polynomial optimization. These include the questions of (i) finding a local minimum, (ii) testing local minimality of a point, and (iii) deciding attainment of the optimal value. Our results characterize the complexity of these three questions for all degrees of the defining polynomials left open by prior literature. Regarding (i) and (ii), we show that unless P=NP, there cannot be a polynomial-time algorithm that finds a point within Euclidean distance $c^n$ (for any constant $c$) of a local minimum of an $n$-variate quadratic program. By contrast, we show that a local minimum of a cubic polynomial can be found efficiently by semidefinite programming (SDP). We prove that second-order points of cubic polynomials admit an efficient semidefinite representation, even though their critical points are NP-hard to find. We also give an efficiently-checkable necessary and sufficient condition for local minimality of a point for a cubic polynomial. Regarding (iii), we prove that testing whether a quadratically constrained quadratic program with a finite optimal value has an optimal solution is NP-hard. We also show that testing coercivity of the objective function, compactness of the feasible set, and the Archimedean property associated with the description of the feasible set are all NP-hard. We also give a new characterization of coercive polynomials that lends itself to a hierarchy of SDPs. In our final chapter, we present an SDP relaxation for finding approximate Nash equilibria in bimatrix games. We show that for a symmetric game, a $1/3$-Nash equilibrium can be efficiently recovered from any rank-2 solution to this relaxation. We also propose SDP relaxations for NP-hard problems related to Nash equilibria, such as that of finding the highest achievable welfare under any Nash equilibrium.

Amir Ali Ahmadi, Jeffrey Zhang

We consider the notions of (i) critical points, (ii) second-order points, (iii) local minima, and (iv) strict local minima for multivariate polynomials. For each type of point, and as a function of the degree of the polynomial, we study the complexity of deciding (1) if a given point is of that type, and (2) if a polynomial has a point of that type. Our results characterize the complexity of these two questions for all degrees left open by prior literature. Our main contributions reveal that many of these questions turn out to be tractable for cubic polynomials. In particular, we present an efficiently-checkable necessary and sufficient condition for local minimality of a point for a cubic polynomial. We also show that a local minimum of a cubic polynomial can be efficiently found by solving semidefinite programs of size linear in the number of variables. By contrast, we show that it is strongly NP-hard to decide if a cubic polynomial has a critical point. We also prove that the set of second-order points of any cubic polynomial is a spectrahedron, and conversely that any spectrahedron is the projection of the set of second-order points of a cubic polynomial. In our final section, we briefly present a potential application of finding local minima of cubic polynomials to the design of a third-order Newton method.

Amir Ali Ahmadi, Jeffrey Zhang

We show that unless P=NP, there cannot be a polynomial-time algorithm that finds a point within Euclidean distance $c^n$ (for any constant $c \ge 0$) of a local minimizer of an $n$-variate quadratic function over a polytope. This result (even with $c=0$) answers a question of Pardalos and Vavasis that appeared in 1992 on a list of seven open problems in complexity theory for numerical optimization. Our proof technique also implies that the problem of deciding whether a quadratic function has a local minimizer over an (unbounded) polyhedron, and that of deciding if a quartic polynomial has a local minimizer are NP-hard.

Ahmet Iscen, Jeffrey Zhang, Svetlana Lazebnik et al.

We introduce an approach for incremental learning that preserves feature descriptors of training images from previously learned classes, instead of the images themselves, unlike most existing work. Keeping the much lower-dimensional feature embeddings of images reduces the memory footprint significantly. We assume that the model is updated incrementally for new classes as new data becomes available sequentially.This requires adapting the previously stored feature vectors to the updated feature space without having access to the corresponding original training images. Feature adaptation is learned with a multi-layer perceptron, which is trained on feature pairs corresponding to the outputs of the original and updated network on a training image. We validate experimentally that such a transformation generalizes well to the features of the previous set of classes, and maps features to a discriminative subspace in the feature space. As a result, the classifier is optimized jointly over new and old classes without requiring old class images. Experimental results show that our method achieves state-of-the-art classification accuracy in incremental learning benchmarks, while having at least an order of magnitude lower memory footprint compared to image-preserving strategies.

Amir Ali Ahmadi, Jeffrey Zhang

We prove that unless P=NP, there exists no polynomial time (or even pseudo-polynomial time) algorithm that can test whether the optimal value of a nonlinear optimization problem where the objective and constraints are given by low-degree polynomials is attained. If the degrees of these polynomials are fixed, our results along with previously-known "Frank-Wolfe type" theorems imply that exactly one of two cases can occur: either the optimal value is attained on every instance, or it is strongly NP-hard to distinguish attainment from non-attainment. We also show that testing for some well-known sufficient conditions for attainment of the optimal value, such as coercivity of the objective function and closedness and boundedness of the feasible set, is strongly NP-hard. As a byproduct, our proofs imply that testing the Archimedean property of a quadratic module is strongly NP-hard, a property that is of independent interest to the convergence of the Lasserre hierarchy. Finally, we give semidefinite programming (SDP)-based sufficient conditions for attainment of the optimal value, in particular a new characterization of coercive polynomials that lends itself to an SDP hierarchy.

Nicholas Egan, Jeffrey Zhang, Kevin Shen

The Generator of a Generative Adversarial Network (GAN) is trained to transform latent vectors drawn from a prior distribution into realistic looking photos. These latent vectors have been shown to encode information about the content of their corresponding images. Projecting input images onto the latent space of a GAN is non-trivial, but previous work has successfully performed this task for latent spaces with a uniform prior. We extend these techniques to latent spaces with a Gaussian prior, and demonstrate our technique's effectiveness.

Geoffrey H. Tison, Jeffrey Zhang, Francesca N. Delling et al.

The electrocardiogram or ECG has been in use for over 100 years and remains the most widely performed diagnostic test to characterize cardiac structure and electrical activity. We hypothesized that parallel advances in computing power, innovations in machine learning algorithms, and availability of large-scale digitized ECG data would enable extending the utility of the ECG beyond its current limitations, while at the same time preserving interpretability, which is fundamental to medical decision-making. We identified 36,186 ECGs from the UCSF database that were 1) in normal sinus rhythm and 2) would enable training of specific models for estimation of cardiac structure or function or detection of disease. We derived a novel model for ECG segmentation using convolutional neural networks (CNN) and Hidden Markov Models (HMM) and evaluated its output by comparing electrical interval estimates to 141,864 measurements from the clinical workflow. We built a 725-element patient-level ECG profile using downsampled segmentation data and trained machine learning models to estimate left ventricular mass, left atrial volume, mitral annulus e' and to detect and track four diseases: pulmonary arterial hypertension (PAH), hypertrophic cardiomyopathy (HCM), cardiac amyloid (CA), and mitral valve prolapse (MVP). CNN-HMM derived ECG segmentation agreed with clinical estimates, with median absolute deviations (MAD) as a fraction of observed value of 0.6% for heart rate and 4% for QT interval. Patient-level ECG profiles enabled quantitative estimates of left ventricular and mitral annulus e' velocity with good discrimination in binary classification models of left ventricular hypertrophy and diastolic function. Models for disease detection ranged from AUROC of 0.94 to 0.77 for MVP. Top-ranked variables for all models included known ECG characteristics along with novel predictors of these traits/diseases.

Jeffrey Zhang, Sravani Gajjala, Pulkit Agrawal et al.

Automated cardiac image interpretation has the potential to transform clinical practice in multiple ways including enabling low-cost serial assessment of cardiac function in the primary care and rural setting. We hypothesized that advances in computer vision could enable building a fully automated, scalable analysis pipeline for echocardiogram (echo) interpretation. Our approach entailed: 1) preprocessing; 2) convolutional neural networks (CNN) for view identification, image segmentation, and phasing of the cardiac cycle; 3) quantification of chamber volumes and left ventricular mass; 4) particle tracking to compute longitudinal strain; and 5) targeted disease detection. CNNs accurately identified views (e.g. 99% for apical 4-chamber) and segmented individual cardiac chambers. Cardiac structure measurements agreed with study report values (e.g. mean absolute deviations (MAD) of 7.7 mL/kg/m2 for left ventricular diastolic volume index, 2918 studies). We computed automated ejection fraction and longitudinal strain measurements (within 2 cohorts), which agreed with commercial software-derived values [for ejection fraction, MAD=5.3%, N=3101 studies; for strain, MAD=1.5% (n=197) and 1.6% (n=110)], and demonstrated applicability to serial monitoring of breast cancer patients for trastuzumab cardiotoxicity. Overall, we found that, compared to manual measurements, automated measurements had superior performance across seven internal consistency metrics with an average increase in the Spearman correlation coefficient of 0.05 (p=0.02). Finally, we developed disease detection algorithms for hypertrophic cardiomyopathy and cardiac amyloidosis, with C-statistics of 0.93 and 0.84, respectively. Our pipeline lays the groundwork for using automated interpretation to support point-of-care handheld cardiac ultrasound and large-scale analysis of the millions of echos archived within healthcare systems.