Juncan Deng
YOU?
Author Swipe
View article: VQ4DiT: Efficient Post-Training Vector Quantization for Diffusion Transformers
VQ4DiT: Efficient Post-Training Vector Quantization for Diffusion Transformers Open
The Diffusion Transformers Models (DiTs) have transitioned the network architecture from traditional UNets to transformers, demonstrating exceptional capabilities in image generation. Although DiTs have been widely applied to high-definiti…
View article: ViM-VQ: Efficient Post-Training Vector Quantization for Visual Mamba
ViM-VQ: Efficient Post-Training Vector Quantization for Visual Mamba Open
Visual Mamba networks (ViMs) extend the selective state space model (Mamba) to various vision tasks and demonstrate significant potential. As a promising compression technique, vector quantization (VQ) decomposes network weights into codeb…
View article: SSVQ: Unleashing the Potential of Vector Quantization with Sign-Splitting
SSVQ: Unleashing the Potential of Vector Quantization with Sign-Splitting Open
Vector Quantization (VQ) has emerged as a prominent weight compression technique, showcasing substantially lower quantization errors than uniform quantization across diverse models, particularly in extreme compression scenarios. However, i…
View article: MVQ: Towards Efficient DNN Compression and Acceleration with Masked Vector Quantization
MVQ: Towards Efficient DNN Compression and Acceleration with Masked Vector Quantization Open
Vector quantization(VQ) is a hardware-friendly DNN compression method that\ncan reduce the storage cost and weight-loading datawidth of hardware\naccelerators. However, conventional VQ techniques lead to significant accuracy\nloss because …
View article: Efficiency Meets Fidelity: A Novel Quantization Framework for Stable Diffusion
Efficiency Meets Fidelity: A Novel Quantization Framework for Stable Diffusion Open
Text-to-image generation of Stable Diffusion models has achieved notable success due to its remarkable generation ability. However, the repetitive denoising process is computationally intensive during inference, which renders Diffusion mod…
View article: VQ4ALL: Efficient Neural Network Representation via a Universal Codebook
VQ4ALL: Efficient Neural Network Representation via a Universal Codebook Open
The rapid growth of the big neural network models puts forward new requirements for lightweight network representation methods. The traditional methods based on model compression have achieved great success, especially VQ technology which …
View article: Efficiency Meets Fidelity: A Novel Quantization Framework for Stable Diffusion
Efficiency Meets Fidelity: A Novel Quantization Framework for Stable Diffusion Open
Text-to-image generation via Stable Diffusion models (SDM) have demonstrated remarkable capabilities. However, their computational intensity, particularly in the iterative denoising process, hinders real-time deployment in latency-sensitiv…
View article: VQ4DiT: Efficient Post-Training Vector Quantization for Diffusion Transformers
VQ4DiT: Efficient Post-Training Vector Quantization for Diffusion Transformers Open
The Diffusion Transformers Models (DiTs) have transitioned the network architecture from traditional UNets to transformers, demonstrating exceptional capabilities in image generation. Although DiTs have been widely applied to high-definiti…
View article: DiffX: Guide Your Layout to Cross-Modal Generative Modeling
DiffX: Guide Your Layout to Cross-Modal Generative Modeling Open
Diffusion models have made significant strides in language-driven and layout-driven image generation. However, most diffusion models are limited to visible RGB image generation. In fact, human perception of the world is enriched by diverse…
View article: Line Segment Matching Fusing Local Gradient Order and Non-Local Structure Information
Line Segment Matching Fusing Local Gradient Order and Non-Local Structure Information Open
Line segment matching is essential for industrial applications such as scene reconstruction, pattern recognition, and VSLAM. To achieve good performance under the scene with illumination changes, we propose a line segment matching method f…
View article: Automated Classification of High-resolution Rock Image Based on Residual Neural Network
Automated Classification of High-resolution Rock Image Based on Residual Neural Network Open
The identification and classification of high-resolution rock images are significant for oil and gas exploration. In recent years, deep learning has been applied in various fields and achieved satisfactory results. This paper presents a ro…
View article: Robust Edge-Direct Visual Odometry based on CNN edge detection and Shi-Tomasi corner optimization
Robust Edge-Direct Visual Odometry based on CNN edge detection and Shi-Tomasi corner optimization Open
In this paper, we propose a robust edge-direct visual odometry (VO) based on CNN edge detection and Shi-Tomasi corner optimization. Four layers of pyramids were extracted from the image in the proposed method to reduce the motion error bet…
View article: Long-Awaited Next-Generation Road Damage Detection and Localization System is Finally Here
Long-Awaited Next-Generation Road Damage Detection and Localization System is Finally Here Open
With the ever-increasing emphasis on road maintenance to a high standard, the need for automated and robust road damage inspection (detection and localization) systems is becoming greater than ever. In this paper, we introduce a realtime r…