Kai Chen

Ruiyuan Gao, Kai Chen, Bo Xiao et al.

The rapid advancement of diffusion models has greatly improved video synthesis, especially in controllable video generation, which is vital for applications like autonomous driving. Although DiT with 3D VAE has become a standard framework for video generation, it introduces challenges in controllable driving video generation, especially for geometry control, rendering existing control methods ineffective. To address these issues, we propose MagicDrive-V2, a novel approach that integrates the MVDiT block and spatial-temporal conditional encoding to enable multi-view video generation and precise geometric control. Additionally, we introduce an efficient method for obtaining contextual descriptions for videos to support diverse textual control, along with a progressive training strategy using mixed video data to enhance training efficiency and generalizability. Consequently, MagicDrive-V2 enables multi-view driving video synthesis with $3.3\times$ resolution and $4\times$ frame count (compared to current SOTA), rich contextual control, and geometric controls. Extensive experiments demonstrate MagicDrive-V2's ability, unlocking broader applications in autonomous driving.

Alan Dolhasz, Chen Ma, Dave Gausebeck et al.

We present a pipeline for generating defurnished replicas of indoor spaces represented as textured meshes and corresponding multi-view panoramic images. To achieve this, we first segment and remove furniture from the mesh representation, extend planes, and fill holes, obtaining a simplified defurnished mesh (SDM). This SDM acts as an ``X-ray'' of the scene's underlying structure, guiding the defurnishing process. We extract Canny edges from depth and normal images rendered from the SDM. We then use these as a guide to remove the furniture from panorama images via ControlNet inpainting. This control signal ensures the availability of global geometric information that may be hidden from a particular panoramic view by the furniture being removed. The inpainted panoramas are used to texture the mesh. We show that our approach produces higher quality assets than methods that rely on neural radiance fields, which tend to produce blurry low-resolution images, or RGB-D inpainting, which is highly susceptible to hallucinations.

Mira Slavcheva, Dave Gausebeck, Kevin Chen et al.

We propose a pipeline that leverages Stable Diffusion to improve inpainting results in the context of defurnishing -- the removal of furniture items from indoor panorama images. Specifically, we illustrate how increased context, domain-specific model fine-tuning, and improved image blending can produce high-fidelity inpaints that are geometrically plausible without needing to rely on room layout estimation. We demonstrate qualitative and quantitative improvements over other furniture removal techniques.

Kevin K. Chen, Anthony C. Gamst, Alden K. Walker

The weights of a neural network are typically initialized at random, and one can think of the functions produced by such a network as having been generated by a prior over some function space. Studying random networks, then, is useful for a Bayesian understanding of the network evolution in early stages of training. In particular, one can investigate why neural networks with huge numbers of parameters do not immediately overfit. We analyze the properties of random scalar-input feed-forward rectified linear unit architectures, which are random linear splines. With weights and biases sampled from certain common distributions, empirical tests show that the number of knots in the spline produced by the network is equal to the number of neurons, to very close approximation. We describe our progress towards a completely analytic explanation of this phenomenon. In particular, we show that random single-layer neural networks are equivalent to integrated random walks with variable step sizes. That each neuron produces one knot on average is equivalent to the associated integrated random walk having one zero crossing on average. We explore how properties of the integrated random walk, including the step sizes and initial conditions, affect the number of crossings. The number of knots in random neural networks can be related to the behavior of extreme learning machines, but it also establishes a prior preventing optimizers from immediately overfitting to noisy training data.

Kevin Chen, Christopher B. Choy, Manolis Savva et al.

We present a method for generating colored 3D shapes from natural language. To this end, we first learn joint embeddings of freeform text descriptions and colored 3D shapes. Our model combines and extends learning by association and metric learning approaches to learn implicit cross-modal connections, and produces a joint representation that captures the many-to-many relations between language and physical properties of 3D shapes such as color and shape. To evaluate our approach, we collect a large dataset of natural language descriptions for physical 3D objects in the ShapeNet dataset. With this learned joint embedding we demonstrate text-to-shape retrieval that outperforms baseline approaches. Using our embeddings with a novel conditional Wasserstein GAN framework, we generate colored 3D shapes from text. Our method is the first to connect natural language text with realistic 3D objects exhibiting rich variations in color, texture, and shape detail. See video at https://youtu.be/zraPvRdl13Q

Kevin K. Chen

Neural networks with rectified linear unit activations are essentially multivariate linear splines. As such, one of many ways to measure the "complexity" or "expressivity" of a neural network is to count the number of knots in the spline model. We study the number of knots in fully-connected feedforward neural networks with rectified linear unit activation functions. We intentionally keep the neural networks very simple, so as to make theoretical analyses more approachable. An induction on the number of layers $l$ reveals a tight upper bound on the number of knots in $\mathbb{R} \to \mathbb{R}^p$ deep neural networks. With $n_i \gg 1$ neurons in layer $i = 1, \dots, l$, the upper bound is approximately $n_1 \dots n_l$. We then show that the exact upper bound is tight, and we demonstrate the upper bound with an example. The purpose of these analyses is to pave a path for understanding the behavior of general $\mathbb{R}^q \to \mathbb{R}^p$ neural networks.

Kevin K. Chen, Anthony Gamst, Alden Walker

ReLU neural networks define piecewise linear functions of their inputs. However, initializing and training a neural network is very different from fitting a linear spline. In this paper, we expand empirically upon previous theoretical work to demonstrate features of trained neural networks. Standard network initialization and training produce networks vastly simpler than a naive parameter count would suggest and can impart odd features to the trained network. However, we also show the forced simplicity is beneficial and, indeed, critical for the wide success of these networks.

Christopher B. Choy, Danfei Xu, JunYoung Gwak et al.

Inspired by the recent success of methods that employ shape priors to achieve robust 3D reconstructions, we propose a novel recurrent neural network architecture that we call the 3D Recurrent Reconstruction Neural Network (3D-R2N2). The network learns a mapping from images of objects to their underlying 3D shapes from a large collection of synthetic data. Our network takes in one or more images of an object instance from arbitrary viewpoints and outputs a reconstruction of the object in the form of a 3D occupancy grid. Unlike most of the previous works, our network does not require any image annotations or object class labels for training or testing. Our extensive experimental analysis shows that our reconstruction framework i) outperforms the state-of-the-art methods for single view reconstruction, and ii) enables the 3D reconstruction of objects in situations when traditional SFM/SLAM methods fail (because of lack of texture and/or wide baseline).