A5064232323- 3D Shape Modeling and Analysis
- Generative Adversarial Networks and Image Synthesis
- Face recognition and analysis
- Advanced Vision and Imaging
- Advanced Aircraft Design and Technologies
- Air Traffic Management and Optimization
- Human Pose and Action Recognition
- Advanced Neural Network Applications
- Domain Adaptation and Few-Shot Learning
- Advanced Image Processing Techniques
- Face and Expression Recognition
- Image Retrieval and Classification Techniques
- Multimodal Machine Learning Applications
- Air Quality and Health Impacts
- Image Processing and 3D Reconstruction
- Video Surveillance and Tracking Methods
- Computer Graphics and Visualization Techniques
- Aviation Industry Analysis and Trends
- COVID-19 diagnosis using AI
- Lung Cancer Diagnosis and Treatment
- Aerospace and Aviation Technology
- Radiomics and Machine Learning in Medical Imaging
- Air Quality Monitoring and Forecasting
- Fetal and Pediatric Neurological Disorders
- Biometric Identification and Security
Google (United States)
2017-2025
Contrastive learning applied to self-supervised representation has seen a resurgence in recent years, leading state of the art performance unsupervised training deep image models. Modern batch contrastive approaches subsume or significantly outperform traditional losses such as triplet, max-margin and N-pairs loss. In this work, we extend approach fully-supervised setting, allowing us effectively leverage label information. Clusters points belonging same class are pulled together embedding...
The goal of this project is to learn a 3D shape representation that enables accurate surface reconstruction, compact storage, efficient computation, consistency for similar shapes, generalization across diverse categories, and inference from depth camera observations. Towards end, we introduce Local Deep Implicit Functions (LDIF), decomposes space into structured set learned implicit functions. We provide networks infer the decomposition local deep functions mesh or posed image. During...
Template 3D shapes are useful for many tasks in graphics and vision, including fitting observation data, analyzing shape collections, transferring attributes. Because of the variety geometry topology real-world shapes, previous methods generally use a library hand-made templates. In this paper, we investigate learning general template from data. To allow widely varying topology, choose an implicit surface representation based on composition local elements. While long known to computer...
We present a method for training regression network from image pixels to 3D morphable model coordinates using only unlabeled photographs. The loss is based on features facial recognition network, computed on-the-fly by rendering the predicted faces with differentiable renderer. To make feasible and avoid fooling effects, we introduce three objectives: batch distribution that encourages output match of model, loopback ensures can correctly reinterpret its own output, multi-view identity...
We present a method for synthesizing frontal, neutral-expression image of person's face given an input photograph. This is achieved by learning to generate facial landmarks and textures from features extracted facial-recognition network. Unlike previous generative approaches, our encoding feature vector largely invariant lighting, pose, expression. Exploiting this invariance, we train decoder network using only photographs. Since these photographs are well aligned, can decompose them into...
Image extension models have broad applications in image editing, computational photography and computer graphics. While inpainting has been extensively studied the literature, it is challenging to directly apply state-of-the-art methods as they tend generate blurry or repetitive pixels with inconsistent semantics. We introduce semantic conditioning discriminator of a generative adversarial network (GAN), achieve strong results on coherent semantics visually pleasing colors textures. also...
Background Developing deep learning models for radiology requires large data sets and substantial computational resources. Data set size limitations can be further exacerbated by distribution shifts, such as rapid changes in patient populations standard of care during the COVID-19 pandemic. A common partial mitigation is transfer pretraining a "generic network" on nonmedical then fine-tuning task-specific set. Purpose To reduce requirements chest radiography using an advanced machine...
Abstract Persistent contrails make up a large fraction of aviation's contribution to global warming. We describe scalable, automated detection and matching (ADM) system determine from satellite data whether flight has made persistent contrail. The ADM compares segments detected by computer vision algorithm running on images the GOES-16 Advanced Baseline Imager. develop use it label each segment as match or non-match. perform this analysis 1.6 million segments. result is an which flights...
Abstract. Contrail cirrus clouds persisting in ice-supersaturated air cause a substantial fraction of aviation's climate impact. One proposed method for the mitigation this impact involves modifying flight paths to avoid particular regions atmosphere that are conducive formation persistent contrails. Ascertaining which formed each observed contrail can be used assess and improve forecast models, as well study effectiveness performing avoidance. The problem contrail-to-flight attribution is...
Contrails (condensation trails) are line-shaped ice clouds caused by aircraft and likely the largest contributor of aviation-induced climate change. Contrail avoidance is potentially an inexpensive way to significantly reduce impact aviation. An automated contrail detection system essential tool develop evaluate systems. In this paper, we present a human-labeled dataset named OpenContrails train models based on GOES-16 Advanced Baseline Imager (ABI) data. We propose model that incorporates...
We introduce CAN, a simple, efficient and scalable method for self-supervised learning of visual representations. Our framework is minimal conceptually clean synthesis (C) contrastive learning, (A) masked autoencoders, (N) the noise prediction approach used in diffusion models. The mechanisms are complementary to one another: shapes embedding space across batch image samples; autoencoders focus on reconstruction low-frequency spatial correlations single sample; encourages high-frequency...
Contrails (condensation trails) are line-shaped ice clouds caused by aircraft and a substantial contributor to aviation-induced climate change. Contrail avoidance is potentially an inexpensive way significantly reduce the impact of aviation. An automated contrail detection system essential tool develop evaluate systems. In this article, we present human-labeled dataset named OpenContrails train models based on GOES-16 Advanced Baseline Imager (ABI) data. We propose model that incorporates...
Abstract Contrails, formed by aircraft engines, are a major component of aviation’s impact on anthropogenic climate change. Contrail avoidance is potential option to mitigate this warming effect, however, uncertainties surrounding operational constraints and accurate formation prediction make it unclear whether feasible. Here we address gap with feasibility test through randomized controlled trial contrail in commercial aviation at the per-flight level. Predictions for regions prone came...