lpips
PulseAugur coverage of lpips — every cluster mentioning lpips across labs, papers, and developer communities, ranked by signal.
- instance of peak signal-to-noise ratio 90%
- used by peak signal-to-noise ratio 70%
- used by Structural Similarity Index Measure 70%
- instance of Structural Similarity Index Measure 70%
- used by Gotit.pub 70%
- competes with Structural Similarity Index Measure 60%
- uses Structural Similarity Index Measure 60%
9 day(s) with sentiment data
-
New video coding method balances perceptual quality and temporal consistency
Researchers have developed a new framework for neural video coding that balances perceptual quality with reconstruction fidelity. This approach uses a local sliced Wasserstein distortion (SWD) to match feature distribut…
-
New PRIN network enhances low-light images with perceptual and structural focus
Researchers have developed a new network called PRIN (Perceptual-Structural Retinex-Inspired Network) designed to enhance low-light images. This method decomposes images into reflectance and illumination components, app…
-
New PBSR method enhances single-image resolution using synthesized pseudo-burst frames
Researchers have developed a novel method called Pseudo-Burst Super-Resolution (PBSR) that enhances image resolution from a single input by synthesizing a pseudo-burst of frames. This technique utilizes zero-shot novel …
-
New ReMATF framework efficiently mitigates video turbulence
Researchers have developed ReMATF, a novel recurrent framework designed to mitigate atmospheric turbulence in videos. This lightweight system processes only two frames at a time, significantly reducing computational cos…
-
ParetoGS advances 3D Gaussian Splatting with continuous budget optimization
Researchers have introduced ParetoGS, a novel method for 3D Gaussian Splatting that optimizes scene representations for varying Gaussian counts. Unlike previous approaches that may require separate optimization for each…
-
ResGVC framework enhances generative video compression with residual coding
Researchers have introduced ResGVC, a novel framework for generative video compression that enhances fidelity by transmitting residual signals. This method guides denoising towards original images by encoding residual s…
-
Loomis Painter framework reconstructs artistic painting processes
Researchers have developed Loomis Painter, a novel framework designed to reconstruct and understand the process of artistic painting. This system tackles the challenge of generating temporally and structurally consisten…
-
SemQAM advances generative image transmission with joint coding and modulation
Researchers have developed SemQAM, a novel scheme for joint semantic-channel coding and modulation designed to improve generative image transmission. Unlike traditional methods that optimize these components separately,…
-
Deep learning assesses road lane quality from dashcam video
Researchers have developed a new method for assessing road lane quality using deep learning and dashcam video. The system utilizes a geo-referenced dataset of over 63,000 frames collected in South Texas. It introduces t…
-
New 3D Gaussian Splatting Method Improves Scene Densification with Occlusion Awareness
Researchers have developed a new method for 3D Gaussian Splatting that improves scene densification by accounting for occlusion. This occlusion-aware approach ensures that density statistics are only gathered from views…
-
Brain-Inspired Engram Memory Network Enhances AI Explainability
Researchers have developed an Engram Memory Network (EMN), a novel AI model inspired by the human memory system, to improve the explainability and reliability of prototype-based AI. The EMN utilizes a Hopfield-inspired …
-
CALM method optimizes video generation by calibrating KV-cache compression
Researchers have developed CALM (Calibrated Allocation for Layer-Head Merging), a novel method to reduce KV-cache memory usage in autoregressive video generation. Unlike previous approaches that applied uniform merging …
-
InfiniSplat v2 improves 3D Gaussian reconstruction for novel-view rendering · 2 sources tracked
InfiniSplat v2 is a new method for feed-forward 3D Gaussian reconstruction that enhances novel-view rendering quality. It achieves this by decoupling geometry and attribute prediction, aligning Gaussians with ground-tru…
-
New 3DGS method improves reconstruction quality using error-driven densification
Researchers have developed a new confidence-aware densification method for 3D Gaussian Splatting (3DGS) that improves reconstruction quality, particularly at lower primitive budgets. This approach prioritizes placing ne…
-
New framework sets privacy-sensitive reconstruction bounds for DP-SGD models
Researchers have developed a new information-theoretic framework to establish reconstruction bounds for models trained with Differential Privacy (DP-SGD). This framework provides a distortion bound that can be calculate…
-
New AI pipeline simulates abdominal ultrasounds from semantic labels
Researchers have developed a novel two-stage pipeline for simulating abdominal ultrasound images from semantic labels, bypassing the need for patient-specific CT scans during inference. The first stage uses models like …
-
New Kuration SDK aims to bridge virtual-to-real gap in AI world models
Researchers have developed the Kuration SDK, a new toolkit designed to bridge the gap between simulated and real-world performance in action-conditioned world models. The SDK addresses limitations in current benchmarks,…
-
New SRIM method enhances image protection against malicious edits
Researchers have developed a new method called SRIM to protect images from malicious editing, particularly when the images are resized. Existing methods are optimized for specific resolutions, but SRIM accounts for unkn…
-
New video codec uses event streams to improve RGB compression efficiency
Researchers have developed an Event-guided Neural Video Codec (ENVC) that leverages event streams, which capture brightness changes between frames, to enhance the efficiency of RGB video compression. The ENVC model inte…
-
New Jacobian Lens Distance method improves image and video perception analysis
Researchers have introduced the Jacobian Lens Distance (JLD), a novel method for measuring perceptual differences between images. Unlike previous methods that rely on human labels or struggle with resolution changes, JL…