While visual question-answering (VQA) benchmarks have catalyzed the development of reasoning techniques, they have focused on vertical thinking. Effective problem-solving also necessitates lateral thinking, which remains understudied in AI and has not been used to test visual perception systems. To bridge this gap, we formulate visual lateral thinking as a multiple-choice question-answering task and describe a three-step taxonomy-driven methodology for instantiating task examples. Then, we develop COLUMBUS, a synthetic benchmark that applies the task pipeline to create QA sets with text and icon rebus puzzles based on publicly available collections of compounds and common phrases. COLUMBUS comprises over 1,000 puzzles, each with four answer candidates. While the SotA vision language models (VLMs) achieve decent performance, our evaluation demonstrates a substantial gap between humans and models. VLMs benefit from human-curated descriptions but struggle to self-generate such representations at the right level of abstraction.

In recent years there have been remarkable breakthroughs in image-to-video generation. However, the 3D consistency and camera controllability of generated frames have remained unsolved. Recent studies have attempted to incorporate camera control into the generation process, but their results are often limited to simple trajectories or lack the ability to generate consistent videos from multiple distinct camera paths for the same scene. To address these limitations, we introduce Cavia, a novel framework for camera-controllable, multi-view video generation, capable of converting an input image into multiple spatiotemporally consistent videos. Our framework extends the spatial and temporal attention modules into view-integrated attention modules, improving both viewpoint and temporal consistency. This flexible design allows for joint training with diverse curated data sources, including scene-level static videos, object-level synthetic multi-view dynamic videos, and real-world monocular dynamic videos. To the best of our knowledge, Cavia is the first framework that enables users to generate multiple videos of the same scene with precise control over camera motion, while simultaneously preserving object motion. Extensive experiments demonstrate that Cavia surpasses state-of-the-art methods in terms of geometric consistency and perceptual quality.

Generative image editing has recently witnessed extremely fast-paced growth. Some works use high-level conditioning such as text, while others use low-level conditioning. Nevertheless, most of them lack fine-grained control over the properties of the different objects present in the image, i.e. object-level image editing. In this work, we tackle the task by perceiving the images as an amalgamation of various objects and aim to control the properties of each object in a fine-grained manner. Out of these properties, we identify structure and appearance as the most intuitive to understand and useful for editing purposes. We propose PAIR Diffusion, a generic framework that enables a diffusion model to control the structure and appearance properties of each object in the image. We show that having control over the properties of each object in an image leads to comprehensive editing capabilities. Our framework allows for various object-level editing operations on real images such as reference image-based appearance editing, free-form shape editing, adding objects, and variations. Thanks to our design, we do not require any inversion step. Additionally, we propose multi-modal classifier-free guidance which enables editing images using both reference images and text when using our approach with foundational diffusion models. We validate the above claims by extensively evaluating our framework on both unconditional and foundational diffusion models.

The task of novel view synthesis aims to generate unseen perspectives of an object or scene from a limited set of input images. Nevertheless, synthesizing novel views from a single image remains a significant challenge. Previous approaches tackle this problem by adopting mesh prediction, multi-plane image construction, or more advanced techniques such as neural radiance fields. Recently, a pre-trained diffusion model that is specifically designed for 2D image synthesis has demonstrated its capability in producing photorealistic novel views, if sufficiently optimized with a 3D finetuning task. Despite greatly improved fidelity and generalizability, training such a powerful diffusion model requires a vast volume of training data and model parameters, resulting in a notoriously long time and high computational costs. To tackle this issue, we propose Efficient-3DiM, a highly efficient yet effective framework to learn a single-image novel-view synthesizer. Motivated by our in-depth analysis of the diffusion model inference process, we propose several pragmatic strategies to reduce training overhead to a manageable scale, including a crafted timestep sampling strategy, a superior 3D feature extractor, and an enhanced training scheme. When combined, our framework can reduce the total training time from 10 days to less than 1 day, significantly accelerating the training process on the same computational platform (an instance with 8 Nvidia A100 GPUs). Comprehensive experiments are conducted to demonstrate the efficiency and generalizability of our proposed method.

While multi-modal large language models (MLLMs) have shown significant progress across popular visual reasoning benchmarks, whether they possess abstract visual reasoning abilities remains an open question. Similar to the Sudoku puzzles, abstract visual reasoning (AVR) problems require finding high-level patterns (e.g., repetition constraints on numbers) that control the input shapes (e.g., digits) in a specific task configuration (e.g., matrix). However, existing AVR benchmarks only consider a limited set of patterns (addition, conjunction), input shapes (rectangle, square), and task configurations (3 × 3 matrices). And they fail to capture all abstract reasoning patterns in human cognition necessary for addressing real-world tasks, such as geometric properties and object boundary understanding in real-world navigation. To evaluate MLLMs’ AVR abilities systematically, we introduce MARVEL founded on the core knowledge system in human cognition, a multi-dimensional AVR benchmark with 770 puzzles composed of six core knowledge patterns, geometric and abstract shapes, and five different task configurations. To inspect whether the model performance is grounded in perception or reasoning, MARVEL complements the standard AVR question with perception questions in a hierarchical evaluation framework. We conduct comprehensive experiments on MARVEL with ten representative MLLMs in zero-shot and few-shot settings. Our experiments reveal that all MLLMs show near-random performance on MARVEL, with significant performance gaps (40%) compared to humans across all patterns and task configurations. Further analysis of perception questions reveals that MLLMs struggle to comprehend the visual features (near-random performance). Although closed-source MLLMs, such as GPT-4V, show a promising understanding of reasoning patterns (on par with humans) after adding textual descriptions, this advantage is hindered by their weak perception abilities. We release our entirecode and dataset at https://github.com/1171-jpg/MARVEL_AVR.

Food computing has emerged as a prominent multidisciplinary field of research in recent years. An ambitious goal of food computing is to develop end-to-end intelligent systems capable of autonomously producing recipe information for a food image. Current image-to-recipe methods are retrieval-based and their success depends heavily on the dataset size and diversity, as well as the quality of learned embeddings. Meanwhile, the emergence of powerful attention-based vision and language models presents a promising avenue for accurate and generalizable recipe generation, which has yet to be extensively explored. This paper proposes FIRE, a novel multimodal methodology tailored to recipe generation in the food computing domain, which generates the food title, ingredients, and cooking instructions based on input food images. FIRE leverages the BLIP model to generate titles, utilizes a Vision Transformer with a decoder for ingredient extraction, and employs the T5 model to generate recipes incorporating titles and ingredients as inputs. We showcase two practical applications that can benefit from integrating FIRE with large language model prompting: recipe customization to fit recipes to user preferences and recipe-to-code transformation to enable automated cooking processes. Our experimental findings validate the efficacy of our proposed approach, underscoring its potential for future advancements and widespread adoption in food computing.

Abstract As a core cognitive skill that enables the transferability of information across domains, analogical reasoning has been extensively studied for both humans and computational models. However, while cognitive theories of analogy often focus on narratives and study the distinction between surface, relational, and system similarities, existing work in natural language processing has a narrower focus as far as relational analogies between word pairs. This gap brings a natural question: can state-of-the-art large language models (LLMs) detect system analogies between narratives? To gain insight into this question and extend word-based relational analogies to relational system analogies, we devise a comprehensive computational framework that operationalizes dominant theories of analogy, using narrative elements to create surface and system mappings. Leveraging the interplay between these mappings, we create a binary task and benchmark for Analogical Reasoning on Narratives (ARN), covering four categories of far (cross-domain)/near (within-domain) analogies and disanalogies. We show that while all LLMs can largely recognize near analogies, even the largest ones struggle with far analogies in a zero-shot setting, with GPT4.0 scoring below random. Guiding the models through solved examples and Chain-of-Thought reasoning enhances their analogical reasoning ability. Yet, since even in the few-shot setting, the best model only performs halfway between random and humans, ARN opens exciting directions for computational analogical reasoners.

Neural Radiance Fields (NeRFs) are a powerful representation for modeling a 3D scene as a continuous function. Though NeRF is able to render complex 3D scenes with view-dependent effects, few efforts have been devoted to exploring its limits in a high-resolution setting. Specifically, existing NeRF-based methods face several limitations when reconstructing high-resolution real scenes, including a very large number of parameters, misaligned input data, and overly smooth details. In this work, we conduct the first pilot study on training NeRF with high-resolution data and propose the corresponding solutions: 1) marrying the multilayer perceptron (MLP) with convolutional layers which can encode more neighborhood information while reducing the total number of parameters; 2) a novel training strategy to address misalignment caused by moving objects or small camera calibration errors; and 3) a high-frequency aware loss. Our approach is nearly free without introducing obvious training/testing costs, while experiments on different datasets demonstrate that it can recover more high-frequency details compared with the current state-of-the-art NeRF models. Project page: https://yifanjiang19.github.io/alignerf.

Virtual reality and augmented reality (XR) bring increasing demand for 3D content generation. However, creating high-quality 3D content requires tedious work from a human expert. In this work, we study the challenging task of lifting a single image to a 3D object and, for the first time, demonstrate the ability to generate a plausible 3D object with 360° views that corresponds well with the given reference image. By conditioning on the reference image, our model can fulfill the everlasting curiosity for synthesizing novel views of objects from images. Our technique sheds light on a promising direction of easing the workflows for 3D artists and XR designers. We propose a novel framework, dubbed NeuralLift-360, that utilizes a depth-aware neural radiance representation (NeRF) and learns to craft the scene guided by denoising diffusion models. By introducing a ranking loss, our NeuralLift-360 can be guided with rough depth estimation in the wild. We also adopt a CLIP-guided sampling strategy for the diffusion prior to provide coherent guidance. Extensive experiments demonstrate that our NeuralLift-360 significantly outperforms existing state-of-the-art baselines. Project page: https://vita-group.github.io/NeuralLift-360/

Stories about everyday situations are an essential part of human communication, motivating the need to develop AI agents that can reliably understand these stories. Despite the long list of supervised methods for story completion and procedural understanding, current AI fails to generalize its procedural reasoning to unseen stories. This paper is based on the hypothesis that the generalization can be improved by associating downstream prediction with fine-grained modeling and the abstraction of procedural knowledge in stories. To test this hypothesis, we design LEAP: a comprehensive framework that reasons over stories by jointly considering their (1) overall plausibility, (2) conflict sentence pairs, and (3) participant physical states. LEAP integrates state-of-the-art modeling architectures, training regimes, and augmentation strategies based on natural and synthetic stories. To address the lack of densely annotated training data on participants and their physical states, we devise a robust automatic labeler based on semantic parsing and few-shot prompting with large language models. Our experiments with in- and out-of-domain tasks reveal insights into the interplay of architectures, training regimes, and augmentation strategies. LEAP’s labeler consistently improves performance on out-of-domain datasets, while our case studies show that the dense annotation supports explainability.

The success of language models has inspired the NLP community to attend to tasks that require implicit and complex reasoning, relying on human-like commonsense mechanisms.While such vertical thinking tasks have been relatively popular, lateral thinking puzzles have received little attention.To bridge this gap, we devise BRAINTEASER: a multiple-choice Question Answering task designed to test the model's ability to exhibit lateral thinking and defy default commonsense associations.We design a three-step procedure for creating the first lateral thinking benchmark, consisting of data collection, distractor generation, and generation of reconstruction examples, leading to 1,100 puzzles with high-quality annotations.To assess the consistency of lateral reasoning by models, we enrich BRAINTEASER based on a semantic and contextual reconstruction of its questions.Our experiments with state-ofthe-art instruction-and commonsense language models reveal a significant gap between human and model performance, which is further widened when consistency across reconstruction formats is considered.We make all of our code and data available to stimulate work on developing and evaluating lateral thinking models.

We present Prompt Diffusion, a framework for enabling in-context learning in diffusion-based generative models. Given a pair of task-specific example images, such as depth from/to image and scribble from/to image, and a text guidance, our model automatically understands the underlying task and performs the same task on a new query image following the text guidance. To achieve this, we propose a vision-language prompt that can model a wide range of vision-language tasks and a diffusion model that takes it as input. The diffusion model is trained jointly on six different tasks using these prompts. The resulting Prompt Diffusion model becomes the first diffusion-based vision-language foundation model capable of in-context learning. It demonstrates high-quality in-context generation for the trained tasks and effectively generalizes to new, unseen vision tasks using their respective prompts. Our model also shows compelling text-guided image editing results. Our framework aims to facilitate research into in-context learning for computer vision. We share our code and pre-trained models at https://github.com/Zhendong-Wang/Prompt-Diffusion.

Diffusion models are powerful, but they require a lot of time and data to train. We propose Patch Diffusion, a generic patch-wise training framework, to significantly reduce the training time costs while improving data efficiency, which thus helps democratize diffusion model training to broader users. At the core of our innovations is a new conditional score function at the patch level, where the patch location in the original image is included as additional coordinate channels, while the patch size is randomized and diversified throughout training to encode the cross-region dependency at multiple scales. Sampling with our method is as easy as in the original diffusion model. Through Patch Diffusion, we could achieve $\mathbf{\ge 2\times}$ faster training, while maintaining comparable or better generation quality. Patch Diffusion meanwhile improves the performance of diffusion models trained on relatively small datasets, $e.g.$, as few as 5,000 images to train from scratch. We achieve outstanding FID scores in line with state-of-the-art benchmarks: 1.77 on CelebA-64$\times$64, 1.93 on AFHQv2-Wild-64$\times$64, and 2.72 on ImageNet-256$\times$256. We share our code and pre-trained models at https://github.com/Zhendong-Wang/Patch-Diffusion.

In blurry images, the degree of image blurs may vary drastically due to different factors, such as varying speeds of shaking cameras and moving objects, as well as defects of the camera lens. However, current end-to-end models failed to explicitly take into account such diversity of blurs. This unawareness compromises the specialization at each blur level, yielding sub-optimal deblurred images as well as redundant post-processing. Therefore, how to specialize one model simultaneously at different blur levels, while still ensuring coverage and generalization, becomes an emerging challenge. In this work, we propose Ada-Deblur, a super-network that can be applied to a "broad spectrum" of blur levels with no re-training on novel blurs. To balance between individual blur level specialization and wide-range blur levels coverage, the key idea is to dynamically adapt the network architectures from a single well-trained super-network structure, targeting flexible image processing with different deblurring capacities at test time. Extensive experiments demonstrate that our work outperforms strong baselines by demonstrating better reconstruction accuracy while incurring minimal computational overhead. Besides, we show that our method is effective for both synthetic and realistic blurs compared to these baselines. The performance gap between our model and the state-of-the-art becomes more prominent when testing with unseen and strong blur levels. Specifically, our model demonstrates surprising deblurring performance on these images with PSNR improvements of around 1 dB. Our code is publicly available at https://github.com/wuqiuche/Ada-Deblur.

Low-light enhancement is an increasingly important function in image editing and visual creation. Most existing enhancing algorithms are trained to enlighten a given image in a globally homogeneous way, and (implicitly) to some predefined extent of brightness. They are neither capable of enhancing only local regions of interest ("where") while keeping the overall visual appearance plausible, nor producing outputs at a range of different illumination levels ("how much"). Those hurdles significantly limit the prospect of flexible, customizable, or even user-interactive low-light enhancement. To address these gaps, we propose Region-Controllable Robust Light Enhancement (ReCoRo), a novel framework that allows users to directly specify "where" and "how much" they want to enhance from an input low-light image; meanwhile, the model will learn to intelligently maintain the overall consistent visual appearance and plausible composition via a discriminator. Moreover, since in practical mobile APPs, such user specifications often come in imprecise forms (e.g., finger-drawn masks), we propose to bake in domain-specific data augmentations into training ReCoRo, so that the learned model can gain resilience to various roughly-supplied user masks. Up to our best knowledge, ReCoRo is the first of its kind that allows the user to localize the enlightenment region as well as to control the light intensity. Extensive experiments clearly demonstrate that ReCoRo outperforms state-of-the-art methods in terms of qualitative results, quantitative metrics, and versatile controllability. Project repository: https://bit.ly/ReCoRo-lowlight.

Implicit Neural Representations (INRs) encoding continuous multi-media data via multi-layer perceptrons has shown undebatable promise in various computer vision tasks. Despite many successful applications, editing and processing an INR remains intractable as signals are represented by latent parameters of a neural network. Existing works manipulate such continuous representations via processing on their discretized instance, which breaks down the compactness and continuous nature of INR. In this work, we present a pilot study on the question: how to directly modify an INR without explicit decoding? We answer this question by proposing an implicit neural signal processing network, dubbed INSP-Net, via differential operators on INR. Our key insight is that spatial gradients of neural networks can be computed analytically and are invariant to translation, while mathematically we show that any continuous convolution filter can be uniformly approximated by a linear combination of high-order differential operators. With these two knobs, INSP-Net instantiates the signal processing operator as a weighted composition of computational graphs corresponding to the high-order derivatives of INRs, where the weighting parameters can be data-driven learned. Based on our proposed INSP-Net, we further build the first Convolutional Neural Network (CNN) that implicitly runs on INRs, named INSP-ConvNet. Our experiments validate the expressiveness of INSP-Net and INSP-ConvNet in fitting low-level image and geometry processing kernels (e.g. blurring, deblurring, denoising, inpainting, and smoothening) as well as for high-level tasks on implicit fields such as image classification.

Efficient video architecture is the key to deploying video recognition systems on devices with limited computing resources. Unfortunately, existing video architectures are often computationally intensive and not suitable for such applications. The recent X3D work presents a new family of efficient video models by expanding a hand-crafted image architecture along multiple axes, such as space, time, width, and depth. Although operating in a conceptually large space, X3D searches one axis at a time, and merely explored a small set of 30 architectures in total, which does not sufficiently explore the space. This paper bypasses existing 2D architectures, and directly searched for 3D architectures in a fine-grained space, where block type, filter number, expansion ratio and attention block are jointly searched. A probabilistic neural architecture search method is adopted to efficiently search in such a large space. Evaluations on Kinetics and Something-Something-V2 benchmarks confirm our AutoX3D models outperform existing ones in accuracy up to 1.3% under similar FLOPs, and reduce the computational cost up to ×1.74 when reaching similar performance.

Low-light images captured in the real world are inevitably corrupted by sensor noise. Such noise is spatially variant and highly dependent on the underlying pixel intensity, deviating from the oversimplified assumptions in conventional denoising. Existing light enhancement methods either overlook the important impact of real-world noise during enhancement, or treat noise removal as a separate pre- or post-processing step. We present Coordinated Enhancement for Real-world Low-light Noisy Images (CERL), that seamlessly integrates light enhancement and noise suppression parts into a unified and physics-grounded optimization framework. For the real low-light noise removal part, we customize a self-supervised denoising model that can easily be adapted without referring to clean ground-truth images. For the light enhancement part, we also improve the design of a state-of-the-art backbone. The two parts are then joint formulated into one principled plug-and-play optimization. Our approach is compared against state-of-the-art low-light enhancement methods both qualitatively and quantitatively. Besides standard benchmarks, we further collect and test on a new realistic low-light mobile photography dataset (RLMP), whose mobile-captured photos display heavier realistic noise than those taken by high-quality cameras. CERL consistently produces the most visually pleasing and artifact-free results across all experiments. Our RLMP dataset and codes are available at: https://github.com/VITA-Group/CERL.

The promise of Deep Neural Network (DNN) powered Internet of Thing (IoT) devices has motivated a tremendous demand for automated solutions to enable fast development and deployment of efficient (1) DNNs equipped with instantaneous accuracy-efficiency trade-off capability to accommodate the time-varying resources at IoT devices and (2) dataflows to optimize DNNs’ execution efficiency on different devices. Therefore, we propose InstantNet to automatically generate and deploy instantaneously switchable-precision networks which operates at variable bit-widths. Extensive experiments show that the proposed InstantNet consistently outperforms state-of-the-art designs. Our codes are available at: https://github.com/RICE-EIC/InstantNet.

Image harmonization aims to improve the quality of image compositing by matching the "appearance" (e.g., color tone, brightness and contrast) between foreground and background images. However, collecting large-scale annotated datasets for this task requires complex professional retouching. Instead, we propose a novel Self-Supervised Harmonization framework (SSH) that can be trained using just "free" natural images without being edited. We reformulate the image harmonization problem from a representation fusion perspective, which separately processes the foreground and background examples, to address the background occlusion issue. This framework design allows for a dual data augmentation method, where diverse [foreground, background, pseudo GT] triplets can be generated by cropping an image with perturbations using 3D color lookup tables (LUTs). In addition, we build a real-world harmonization dataset as carefully created by expert users, for evaluation and benchmarking purposes. Our results show that the proposed self-supervised method outperforms previous state-of-the-art methods in terms of reference metrics, visual quality, and subject user study. Code and dataset are available at https://github.com/VITA-Group/SSHarmonization.

The recent explosive interest on transformers has suggested their potential to become powerful ``universal" models for computer vision tasks, such as classification, detection, and segmentation. While those attempts mainly study the discriminative models, we explore transformers on some more notoriously difficult vision tasks, e.g., generative adversarial networks (GANs). Our goal is to conduct the first pilot study in building a GAN \textit{completely free of convolutions}, using only pure transformer-based architectures. Our vanilla GAN architecture, dubbed \textbf{TransGAN}, consists of a memory-friendly transformer-based generator that progressively increases feature resolution, and correspondingly a multi-scale discriminator to capture simultaneously semantic contexts and low-level textures. On top of them, we introduce the new module of grid self-attention for alleviating the memory bottleneck further, in order to scale up TransGAN to high-resolution generation. We also develop a unique training recipe including a series of techniques that can mitigate the training instability issues of TransGAN, such as data augmentation, modified normalization, and relative position encoding. Our best architecture achieves highly competitive performance compared to current state-of-the-art GANs using convolutional backbones. Specifically, TransGAN sets \textbf{new state-of-the-art} inception score of 10.43 and FID of 18.28 on STL-10. It also reaches the inception score of 9.02 and FID of 9.26 on CIFAR-10, and 5.28 FID on CelebA $\mathbf{128} \times \mathbf{128}$, respectively: both on par with the current best results and outperforming StyleGAN-V2. When it comes to higher-resolution (e.g. $\mathbf{256} \times \mathbf{256}$) generation tasks, such as on CelebA-HQ and LSUN-Church, TransGAN continues to produce diverse visual examples with high fidelity and impressive texture details. In addition, we dive deep into the transformer-based generation models to understand how their behaviors differ from convolutional ones, by visualizing training dynamics. The code is available at: https://github.com/VITA-Group/TransGAN.

The self-attention-based model, transformer, is recently becoming the leading backbone in the field of computer vision. In spite of the impressive success made by transformers in a variety of vision tasks, it still suffers from heavy computation and intensive memory costs. To address this limitation, this paper presents an Interpretability-Aware REDundancy REDuction framework (IA-RED$^2$). We start by observing a large amount of redundant computation, mainly spent on uncorrelated input patches, and then introduce an interpretable module to dynamically and gracefully drop these redundant patches. This novel framework is then extended to a hierarchical structure, where uncorrelated tokens at different stages are gradually removed, resulting in a considerable shrinkage of computational cost. We include extensive experiments on both image and video tasks, where our method could deliver up to 1.4x speed-up for state-of-the-art models like DeiT and TimeSformer, by only sacrificing less than 0.7% accuracy. More importantly, contrary to other acceleration approaches, our method is inherently interpretable with substantial visual evidence, making vision transformer closer to a more human-understandable architecture while being lighter. We demonstrate that the interpretability that naturally emerged in our framework can outperform the raw attention learned by the original visual transformer, as well as those generated by off-the-shelf interpretation methods, with both qualitative and quantitative results. Project Page: http://people.csail.mit.edu/bpan/ia-red/.

Deep learning-based methods have achieved remarkable success in image restoration and enhancement, but are they still competitive when there is a lack of paired training data? As one such example, this paper explores the low-light image enhancement problem, where in practice it is extremely challenging to simultaneously take a low-light and a normal-light photo of the same visual scene. We propose a highly effective unsupervised generative adversarial network, dubbed EnlightenGAN, that can be trained without low/normal-light image pairs, yet proves to generalize very well on various real-world test images. Instead of supervising the learning using ground truth data, we propose to regularize the unpaired training using the information extracted from the input itself, and benchmark a series of innovations for the low-light image enhancement problem, including a global-local discriminator structure, a self-regularized perceptual loss fusion, and the attention mechanism. Through extensive experiments, our proposed approach outperforms recent methods under a variety of metrics in terms of visual quality and subjective user study. Thanks to the great flexibility brought by unpaired training, EnlightenGAN is demonstrated to be easily adaptable to enhancing real-world images from various domains. Our codes and pre-trained models are available at: https://github.com/VITA-Group/EnlightenGAN.

Neural architecture search (NAS) has witnessed prevailing success in image classification and (very recently) segmentation tasks. In this paper, we present the first preliminary study on introducing the NAS algorithm to generative adversarial networks (GANs), dubbed AutoGAN. The marriage of NAS and GANs faces its unique challenges. We define the search space for the generator architectural variations and use an RNN controller to guide the search, with parameter sharing and dynamic-resetting to accelerate the process. Inception score is adopted as the reward, and a multi-level search strategy is introduced to perform NAS in a progressive way. Experiments validate the effectiveness of AutoGAN on the task of unconditional image generation. Specifically, our discovered architectures achieve highly competitive performance compared to current state-of-the-art hand-crafted GANs, e.g., setting new state-of-the-art FID scores of 12.42 on CIFAR-10, and 31.01 on STL-10, respectively. We also conclude with a discussion of the current limitations and future potential of AutoGAN. The code is available at https://github.com/TAMU-VITA/AutoGAN.