Multimodallarge language models (MLLMs) have emerged as a prominent area of interest within the research community, given their proficiency in handling and reasoning with non-textual data, including images and videos. This study seeks to extend the application of MLLMs to the realm of autonomous driving by introducing DriveGPT4, a novel interpretable end-to-end autonomous driving system based on LLMs. Capable of processing multi-frame video inputs and textual queries, DriveGPT4 facilitates the interpretation of vehicle actions, offers pertinent reasoning, and effectively addresses a diverse range of questions posed by users. Furthermore, DriveGPT4 predicts low-level vehicle control signals in an end-to-end fashion. These advanced capabilities are achieved through the utilization of a bespoke visual instruction tuning dataset, specifically tailored for autonomous driving applications, in conjunction with a mix-finetuning training strategy. DriveGPT4 represents the pioneering effort to leverage LLMs for the development of an interpretable end-to-end autonomous driving solution. Evaluations conducted on the BDD-X dataset showcase the superior qualitative and quantitative performance of DriveGPT4. Additionally, the fine-tuning of domain-specific data enables DriveGPT4 to yield close or even improved results in terms of autonomous driving grounding when contrasted with GPT4-V.

General image super-resolution techniques have difficulties in recovering detailed face structures when applying to low resolution face images. Recent deep learning based methods tailored for face images have achieved improved performance by jointly trained with additional task such as face parsing and landmark prediction. However, multi-task learning requires extra manually labeled data. Besides, most of the existing works can only generate relatively low resolution face images (e.g., 128×128), and their applications are therefore limited. In this paper, we introduce a novel SPatial Attention Residual Network (SPARNet) built on our newly proposed Face Attention Units (FAUs) for face super-resolution. Specifically, we introduce a spatial attention mechanism to the vanilla residual blocks. This enables the convolutional layers to adaptively bootstrap features related to the key face structures and pay less attention to those less feature-rich regions. This makes the training more effective and efficient as the key face structures only account for a very small portion of the face image. Visualization of the attention maps shows that our spatial attention network can capture the key face structures well even for very low resolution faces (e.g., 16×16). Quantitative comparisons on various kinds of metrics (including PSNR, SSIM, identity similarity, and landmark detection) demonstrate the superiority of our method over current state-of-the-arts. We further extend SPARNet with multi-scale discriminators, named as SPARNetHD, to produce high resolution results (i.e., 512×512). We show that SPARNetHD trained with synthetic data cannot only produce high quality and high resolution outputs for synthetically degraded face images, but also show good generalization ability to real world low quality face images.

Inferring cellular trajectories using a variety of omic data is a critical task in single-cell data science. However, accurate prediction of cell fates, and thereby biologically meaningful discovery, is challenged by the sheer size of single-cell data, the diversity of omic data types, and the complexity of their topologies. We present VIA, a scalable trajectory inference algorithm that overcomes these limitations by using lazy-teleporting random walks to accurately reconstruct complex cellular trajectories beyond tree-like pathways (e.g., cyclic or disconnected structures). We show that VIA robustly and efficiently unravels the fine-grained sub-trajectories in a 1.3-million-cell transcriptomic mouse atlas without losing the global connectivity at such a high cell count. We further apply VIA to discovering elusive lineages and less populous cell fates missed by other methods across a variety of data types, including single-cell proteomic, epigenomic, multi-omics datasets, and a new in-house single-cell morphological dataset.

Solitons can self-assemble into stable bound states, also denoted as soliton molecules that exhibit molecule-like dynamics. Soliton molecules have been predominantly investigated in the anomalous dispersion mode-locked fiber lasers. However, the soliton molecule dynamic evolution is still largely unexplored in the normal dispersion regime. We reveal here that, in the normal dispersion regime, the buildup and dissociation dynamics of soliton molecules. Our theoretical prediction indicates that, under different transmission functions of a saturable absorber (SA), a pair of solitons can be formed directly from background noise and then evolve into a soliton molecule through intense repulsive interaction, or a soliton molecule can be triggered to dissociate into a single soliton with transient annihilation and energy transfer. The experimental observation of short-time soliton molecule buildup and a new soliton molecule dissociation process corroborate the theoretical prediction. Furthermore, a long-time soliton molecule buildup ( ∼ 900 m s ) is discovered with single soliton splitting and soliton pair attraction. The buildup time is over four orders of magnitude longer than that of the short-time soliton molecule ( ∼ 21 µ s ). Our work unveils new perspectives into the ultrafast transient process and the interaction dynamics of soliton molecules in complex nonlinear systems.

This paper addresses the problem of photometric stereo, in both calibrated and uncalibrated scenarios, for non-Lambertian surfaces based on deep learning. We first introduce a fully convolutional deep network for calibrated photometric stereo, which we call PS-FCN. Unlike traditional approaches that adopt simplified reflectance models to make the problem tractable, our method directly learns the mapping from reflectance observations to surface normal, and is able to handle surfaces with general and unknown isotropic reflectance. At test time, PS-FCN takes an arbitrary number of images and their associated light directions as input and predicts a surface normal map of the scene in a fast feed-forward pass. To deal with the uncalibrated scenario where light directions are unknown, we introduce a new convolutional network, named LCNet, to estimate light directions from input images. The estimated light directions and the input images are then fed to PS-FCN to determine the surface normals. Our method does not require a pre-defined set of light directions and can handle multiple images in an order-agnostic manner. Thorough evaluation of our approach on both synthetic and real datasets shows that it outperforms state-of-the-art methods in both calibrated and uncalibrated scenarios.