Search

In the field of low-light image enhancement, both traditional Retinex methods and advanced deep learning techniques such as Retinexformer have shown distinct advantages and limitations. Traditional Retinex methods, designed to mimic the human eye's perception of brightness and color, decompose images into illumination and reflection components but struggle with noise management and detail preservation under low light conditions. Retinexformer enhances illumination estimation through traditional self-attention mechanisms, but faces challenges with insufficient interpretability and suboptimal enhancement effects. To overcome these limitations, this paper introduces the RetinexMamba architecture. RetinexMamba not only captures the physical intuitiveness of traditional Retinex methods but also integrates the deep learning framework of Retinexformer, leveraging the computational efficiency of State Space Models (SSMs) to enhance processing speed. This architecture features innovative illumination estimators and damage restorer mechanisms that maintain image quality during enhancement. Moreover, RetinexMamba replaces the IG-MSA (Illumination-Guided Multi-Head Attention) in Retinexformer with a Fused-Attention mechanism, improving the model's interpretability. Experimental evaluations on the LOL dataset show that RetinexMamba outperforms existing deep learning approaches based on Retinex theory in both quantitative and qualitative metrics, confirming its effectiveness and superiority in enhancing low-light images.

Conditional diffusion models can create unseen images in various settings, aiding image interpolation. Interpolation in latent spaces is well-studied, but interpolation with specific conditions like text or poses is less understood. Simple approaches, such as linear interpolation in the space of conditions, often result in images that lack consistency, smoothness, and fidelity. To that end, we introduce a novel training-free technique named Attention Interpolation via Diffusion (AID). Our key contributions include 1) proposing an inner/outer interpolated attention layer; 2) fusing the interpolated attention with self-attention to boost fidelity; and 3) applying beta distribution to selection to increase smoothness. We also present a variant, Prompt-guided Attention Interpolation via Diffusion (PAID), that considers interpolation as a condition-dependent generative process. This method enables the creation of new images with greater consistency, smoothness, and efficiency, and offers control over the exact path of interpolation. Our approach demonstrates effectiveness for conceptual and spatial interpolation. Code and demo are available at https://github.com/QY-H00/attention-interpolation-diffusion., Comment: NeurIPS 2024 Conference Paper

Large language models (LLMs) have demonstrated remarkable proficiency in understanding and generating responses to complex queries through large-scale pre-training. However, the efficacy of these models in memorizing and reasoning among large-scale structured knowledge, especially world knowledge that explicitly covers abundant factual information remains questionable. Addressing this gap, our research investigates whether LLMs can effectively store, recall, and reason with knowledge on a large scale comparable to latest knowledge bases (KBs) such as Wikidata. Specifically, we focus on three crucial aspects to study the viability: (1) the efficiency of LLMs with different sizes in memorizing the exact knowledge in the large-scale KB; (2) the flexibility of recalling the memorized knowledge in response to natural language queries; (3) the capability to infer new knowledge through reasoning. Our findings indicate that while LLMs hold promise as large-scale KBs capable of retrieving and responding with flexibility, enhancements in their reasoning capabilities are necessary to fully realize their potential.

Transition metal dichalcogenides like MoS2 can exist many phases like the semiconducting 2H and the metallic 1T phases which have shown intriguing properties for energy and electrocatalytic applications. However, the 2H and 1T phases normally distribute coherently in a single-layered MoS2 sheet which is accompanied with ubiquitous hetero-phase boundaries. In this work, by using density functional theory and electrochemical measurement, we report strong charge transfer ability between 2H/1T phase boundary of MoS2 and graphene which accounts for a superb coexistence of gravimetric and volumetric capacitances of 272 F g-1 and 685 F cm-3. As a proof-of-concept application, a flexible solid-state asymmetric supercapacitor based on MoS2/graphene is fabricated, showing a remarkable energy and power densities (46.3 mWh cm-3 and 3.013 Wcm-3). Our work shows the promise of promoting the efficiency of charge flow and energy storage through engineering phase boundary and interface in phase-change materials.

Graphene/silicon heterostructures have attracted tremendous interest as a new platform for diverse electronic and photonic devices such as barristors, solar cells, optical modulators, and chemical sensors. The studies to date largely focus on junctions between graphene and lightly-doped silicon, where a Schottky barrier is believed to dominate the carrier transport process. Here we report a systematic investigation of carrier transport across the heterojunctions formed between graphene and highly-doped silicon. By varying the silicon doping level and the measurement temperature, we show that the carrier transport across the graphene/p++-Si heterojunction is dominated by tunneling effect through the native oxide. We further demonstrate that the tunneling current can be effectively modulated by the external gate electrical field, resulting in a vertical tunneling transistor. Benefited from the large density of states of highly doped silicon, our tunneling transistors can deliver a current density over 20 A/cm2, about two orders of magnitude higher than previous graphene/insulator/graphene tunneling transistor at the same on/off ratio.

Methylammonium lead iodide perovskite has attracted intensive interest for its diverse optoelectronic applications. However, most studies to date have been limited to bulk thin films that are difficult to implement for integrated device arrays because of their incompatibility with typical lithography processes. We report the first patterned growth of regular arrays of perovskite microplate crystals for functional electronics and optoelectronics. We show that large arrays of lead iodide microplates can be grown from an aqueous solution through a seeded growth process and can be further intercalated with methylammonium iodide to produce perovskite crystals. Structural and optical characterizations demonstrate that the resulting materials display excellent crystalline quality and optical properties. We further show that perovskite crystals can be selectively grown on prepatterned electrode arrays to create independently addressable photodetector arrays and functional field effect transistors. The ability to grow perovskite microplates and to precisely place them at specific locations offers a new material platform for the fundamental investigation of the electronic and optical properties of perovskite materials and opens a pathway for integrated electronic and optoelectronic systems., Comment: 8 pages, 4 figures

The recently emerged organohalide perovskites (e.g., CH3NH3PbI3) have drawn intense attention for high efficiency solar cells. However, with a considerable solubility in many solvents, these perovskites are not typically compatible with conventional lithography processes for more complicated device fabrications that are important for both fundamental studies and technological applications. Here we report the creation of novel heterojunction devices based on perovskites and two-dimensional (2D) crystals by taking advantage of the layered characteristic of lead iodide (PbI2) and vapor phase intercalation. We show a graphene/perovskite/graphene vertical stack can deliver a highest photoresponsivity of ~950 A/W and photoconductive gain of ~2200, and a graphene/WSe2/perovskite/graphene heterojunction can display a high on/off ratio (~10^6) transistor behavior with distinct gate-tunable diode characteristics and open-circuit voltages. Such unique perovskite-2D heterostructures have significant potential for future optoelectronic research and can enable broad possibilities with compositional tunability of organohalide perovskites and the versatility offered by diverse 2D materials.

The two-dimensional (2D) layered semiconductors such as MoS2 have attracted tremendous interest as a new class of electronic materials. However, there is considerable challenge in making reliable contacts to these atomically thin materials. Here we present a new strategy by using graphene as back electrodes to achieve Ohmic contact to MoS2. With a finite density of states, the Fermi level of graphene can be readily modified by gate potential to ensure a nearly perfect band alignment with MoS2. We demonstrate, for the first time, a transparent contact can be made to MoS2 with essentially zero contact barrier and linear output behaviour at cryogenic temperatures (down to 1.9 K) for both monolayer and multilayer MoS2. Benefiting from the barrier-free transparent contacts, we show that a metal-insulator-transition (MIT) can be observed in a two-terminal MoS2 device, a phenomenon that could be easily masked by Schottky barrier and only seen in four-terminal devices in conventional metal-contacted MoS2 system. With further passivation y born nitride encapsulation, we demonstrate a record high extrinsic (two-terminal) field effect mobility over 1300 cm2/Vs in MoS2.