Search Results for author: Dengxin Dai

Found 105 papers, 52 papers with code

Nighttime Defogging Using High-Low Frequency Decomposition and Grayscale-Color Networks

no code implementations ECCV 2020 Wending Yan, Robby T. Tan, Dengxin Dai

Given an RGB foggy nighttime image, our grayscale module takes the grayscale version of the image as input, and decomposes it into high and low frequency layers.

Vocal Bursts Intensity Prediction

2D Feature Distillation for Weakly- and Semi-Supervised 3D Semantic Segmentation

no code implementations27 Nov 2023 Ozan Unal, Dengxin Dai, Lukas Hoyer, Yigit Baran Can, Luc van Gool

As 3D perception problems grow in popularity and the need for large-scale labeled datasets for LiDAR semantic segmentation increase, new methods arise that aim to reduce the necessity for dense annotations by employing weakly-supervised training.

2D Semantic Segmentation 3D Semantic Segmentation +3

SSB: Simple but Strong Baseline for Boosting Performance of Open-Set Semi-Supervised Learning

1 code implementation ICCV 2023 Yue Fan, Anna Kukleva, Dengxin Dai, Bernt Schiele

In experiments, SSB greatly improves both inlier classification and outlier detection performance, outperforming existing methods by a large margin.

Multi-Task Learning Outlier Detection

Object-centric Cross-modal Feature Distillation for Event-based Object Detection

no code implementations9 Nov 2023 Lei LI, Alexander Liniger, Mario Millhaeusler, Vagia Tsiminaki, Yuanyou Li, Dengxin Dai

In this paper, we develop a novel knowledge distillation approach to shrink the performance gap between these two modalities.

Knowledge Distillation Object +2

U-BEV: Height-aware Bird's-Eye-View Segmentation and Neural Map-based Relocalization

no code implementations20 Oct 2023 Andrea Boscolo Camiletto, Alfredo Bochicchio, Alexander Liniger, Dengxin Dai, Abel Gawel

Efficient relocalization is essential for intelligent vehicles when GPS reception is insufficient or sensor-based localization fails.

Discwise Active Learning for LiDAR Semantic Segmentation

no code implementations23 Sep 2023 Ozan Unal, Dengxin Dai, Ali Tamer Unal, Luc van Gool

Finally we propose a semi-supervised learning approach to utilize all frames within our dataset and improve performance.

Active Learning LIDAR Semantic Segmentation +1

LiDAR Meta Depth Completion

1 code implementation24 Jul 2023 Wolfgang Boettcher, Lukas Hoyer, Ozan Unal, Ke Li, Dengxin Dai

While using a single model, our method yields significantly better results than a non-adaptive baseline trained on different LiDAR patterns.

Depth Completion Monocular Depth Estimation

MTR++: Multi-Agent Motion Prediction with Symmetric Scene Modeling and Guided Intention Querying

1 code implementation30 Jun 2023 Shaoshuai Shi, Li Jiang, Dengxin Dai, Bernt Schiele

Extensive experimental results demonstrate that the MTR framework achieves state-of-the-art performance on the highly-competitive motion prediction benchmarks, while the MTR++ framework surpasses its precursor, exhibiting enhanced performance and efficiency in predicting accurate multimodal future trajectories for multiple agents.

Autonomous Driving motion prediction

HGFormer: Hierarchical Grouping Transformer for Domain Generalized Semantic Segmentation

1 code implementation CVPR 2023 Jian Ding, Nan Xue, Gui-Song Xia, Bernt Schiele, Dengxin Dai

This work studies semantic segmentation under the domain generalization setting, where a model is trained only on the source domain and tested on the unseen target domain.

Domain Generalization Segmentation +1

FreePoint: Unsupervised Point Cloud Instance Segmentation

no code implementations11 May 2023 Zhikai Zhang, Jian Ding, Li Jiang, Dengxin Dai, Gui-Song Xia

Based on the point features, we perform a multicut algorithm to segment point clouds into coarse instance masks as pseudo labels, which are used to train a point cloud instance segmentation model.

Instance Segmentation Segmentation +2

EDAPS: Enhanced Domain-Adaptive Panoptic Segmentation

1 code implementation ICCV 2023 Suman Saha, Lukas Hoyer, Anton Obukhov, Dengxin Dai, Luc van Gool

EDAPS significantly improves the state-of-the-art performance for panoptic segmentation UDA by a large margin of 20% on SYNTHIA-to-Cityscapes and even 72% on the more challenging SYNTHIA-to-Mapillary Vistas.

Domain Adaptation Instance Segmentation +2

Domain Adaptive and Generalizable Network Architectures and Training Strategies for Semantic Image Segmentation

3 code implementations26 Apr 2023 Lukas Hoyer, Dengxin Dai, Luc van Gool

As previous UDA&DG semantic segmentation methods are mostly based on outdated networks, we benchmark more recent architectures, reveal the potential of Transformers, and design the DAFormer network tailored for UDA&DG.

Domain Generalization Image Segmentation +2

Federated Incremental Semantic Segmentation

1 code implementation CVPR 2023 Jiahua Dong, Duzhen Zhang, Yang Cong, Wei Cong, Henghui Ding, Dengxin Dai

Moreover, new clients collecting novel classes may join in the global training of FSS, which further exacerbates catastrophic forgetting.

Federated Learning Relation +2

TrafficBots: Towards World Models for Autonomous Driving Simulation and Motion Prediction

2 code implementations7 Mar 2023 Zhejun Zhang, Alexander Liniger, Dengxin Dai, Fisher Yu, Luc van Gool

We present TrafficBots, a multi-agent policy built upon motion prediction and end-to-end driving, and based on TrafficBots we obtain a world model tailored for the planning module of autonomous vehicles.

Autonomous Driving Model-based Reinforcement Learning +1

Benchmarking the Robustness of LiDAR Semantic Segmentation Models

1 code implementation3 Jan 2023 Xu Yan, Chaoda Zheng, Ying Xue, Zhen Li, Shuguang Cui, Dengxin Dai

In this paper, we aim to comprehensively analyze the robustness of LiDAR semantic segmentation models under various corruptions.

Autonomous Driving Benchmarking +3

Weakly-Supervised Domain Adaptive Semantic Segmentation With Prototypical Contrastive Learning

1 code implementation CVPR 2023 Anurag Das, Yongqin Xian, Dengxin Dai, Bernt Schiele

In this work, we propose a common framework to use different weak labels, e. g. image, point and coarse labels from target domain to reduce this performance gap.

Contrastive Learning Semantic Segmentation +1

Continuous Pseudo-Label Rectified Domain Adaptive Semantic Segmentation With Implicit Neural Representations

no code implementations CVPR 2023 Rui Gong, Qin Wang, Martin Danelljan, Dengxin Dai, Luc van Gool

Unsupervised domain adaptation (UDA) for semantic segmentation aims at improving the model performance on the unlabeled target domain by leveraging a labeled source domain.

Pseudo Label Semantic Segmentation +1

Simulating Road Spray Effects in Automotive Lidar Sensor Models

1 code implementation16 Dec 2022 Clemens Linnhoff, Dominik Scheuble, Mario Bijelic, Lukas Elster, Philipp Rosenberger, Werner Ritter, Dengxin Dai, Hermann Winner

The model conforms to the Open Simulation Interface (OSI) standard and is based on the formation of detection clusters within a spray plume.

object-detection Object Detection

One-Shot Domain Adaptive and Generalizable Semantic Segmentation with Class-Aware Cross-Domain Transformers

no code implementations14 Dec 2022 Rui Gong, Qin Wang, Dengxin Dai, Luc van Gool

Thus, we aim to relieve this need on a large number of real data, and explore the one-shot unsupervised sim-to-real domain adaptation (OSUDA) and generalization (OSDG) problem, where only one real-world data sample is available.

Autonomous Driving Domain Adaptation +1

MIC: Masked Image Consistency for Context-Enhanced Domain Adaptation

1 code implementation CVPR 2023 Lukas Hoyer, Dengxin Dai, Haoran Wang, Luc van Gool

MIC significantly improves the state-of-the-art performance across the different recognition tasks for synthetic-to-real, day-to-nighttime, and clear-to-adverse-weather UDA.

Image Classification object-detection +4

Normalization Perturbation: A Simple Domain Generalization Method for Real-World Domain Shifts

no code implementations8 Nov 2022 Qi Fan, Mattia Segu, Yu-Wing Tai, Fisher Yu, Chi-Keung Tang, Bernt Schiele, Dengxin Dai

Thus, we propose to perturb the channel statistics of source domain features to synthesize various latent styles, so that the trained deep model can perceive diverse potential domains and generalizes well even without observations of target domain data in training.

Autonomous Driving Domain Generalization

Motion Transformer with Global Intention Localization and Local Movement Refinement

2 code implementations27 Sep 2022 Shaoshuai Shi, Li Jiang, Dengxin Dai, Bernt Schiele

Predicting multimodal future behavior of traffic participants is essential for robotic vehicles to make safe decisions.

motion prediction Trajectory Prediction

MTR-A: 1st Place Solution for 2022 Waymo Open Dataset Challenge -- Motion Prediction

1 code implementation20 Sep 2022 Shaoshuai Shi, Li Jiang, Dengxin Dai, Bernt Schiele

In this report, we present the 1st place solution for motion prediction track in 2022 Waymo Open Dataset Challenges.

motion prediction

Ret3D: Rethinking Object Relations for Efficient 3D Object Detection in Driving Scenes

no code implementations18 Aug 2022 Yu-Huan Wu, Da Zhang, Le Zhang, Xin Zhan, Dengxin Dai, Yun Liu, Ming-Ming Cheng

Current efficient LiDAR-based detection frameworks are lacking in exploiting object relations, which naturally present in both spatial and temporal manners.

3D Object Detection Object +2

HRFuser: A Multi-resolution Sensor Fusion Architecture for 2D Object Detection

1 code implementation30 Jun 2022 Tim Broedermann, Christos Sakaridis, Dengxin Dai, Luc van Gool

Besides standard cameras, autonomous vehicles typically include multiple additional sensors, such as lidars and radars, which help acquire richer information for perceiving the content of the driving scene.

Autonomous Vehicles object-detection +3

Deep Gradient Learning for Efficient Camouflaged Object Detection

1 code implementation25 May 2022 Ge-Peng Ji, Deng-Ping Fan, Yu-Cheng Chou, Dengxin Dai, Alexander Liniger, Luc van Gool

This paper introduces DGNet, a novel deep framework that exploits object gradient supervision for camouflaged object detection (COD).

Defect Detection Object +4

Bi-level Alignment for Cross-Domain Crowd Counting

1 code implementation CVPR 2022 Shenjian Gong, Shanshan Zhang, Jian Yang, Dengxin Dai, Bernt Schiele

The main challenge for this task is to achieve high-quality manual annotations on a large amount of training data.

AutoML Crowd Counting +2

HRDA: Context-Aware High-Resolution Domain-Adaptive Semantic Segmentation

1 code implementation27 Apr 2022 Lukas Hoyer, Dengxin Dai, Luc van Gool

Therefore, we propose HRDA, a multi-resolution training approach for UDA, that combines the strengths of small high-resolution crops to preserve fine segmentation details and large low-resolution crops to capture long-range context dependencies with a learned scale attention, while maintaining a manageable GPU memory footprint.

Segmentation Semantic Segmentation +3

LiDAR Snowfall Simulation for Robust 3D Object Detection

1 code implementation CVPR 2022 Martin Hahner, Christos Sakaridis, Mario Bijelic, Felix Heide, Fisher Yu, Dengxin Dai, Luc van Gool

Due to the difficulty of collecting and annotating training data in this setting, we propose a physically based method to simulate the effect of snowfall on real clear-weather LiDAR point clouds.

Autonomous Driving Object +3

Continual Test-Time Domain Adaptation

2 code implementations CVPR 2022 Qin Wang, Olga Fink, Luc van Gool, Dengxin Dai

However, real-world machine perception systems are running in non-stationary and continually changing environments where the target domain distribution can change over time.

Test-time Adaptation

Scribble-Supervised LiDAR Semantic Segmentation

3 code implementations CVPR 2022 Ozan Unal, Dengxin Dai, Luc van Gool

Densely annotating LiDAR point clouds remains too expensive and time-consuming to keep up with the ever growing volume of data.

3D Semantic Segmentation LIDAR Semantic Segmentation +1

Pix2NeRF: Unsupervised Conditional $π$-GAN for Single Image to Neural Radiance Fields Translation

2 code implementations26 Feb 2022 Shengqu Cai, Anton Obukhov, Dengxin Dai, Luc van Gool

We propose a pipeline to generate Neural Radiance Fields~(NeRF) of an object or a scene of a specific class, conditioned on a single input image.

3D-Aware Image Synthesis Novel View Synthesis +2

Adiabatic Quantum Computing for Multi Object Tracking

no code implementations CVPR 2022 Jan-Nico Zaech, Alexander Liniger, Martin Danelljan, Dengxin Dai, Luc van Gool

Multi-Object Tracking (MOT) is most often approached in the tracking-by-detection paradigm, where object detections are associated through time.

Multi-Object Tracking Object

Pix2NeRF: Unsupervised Conditional p-GAN for Single Image to Neural Radiance Fields Translation

1 code implementation CVPR 2022 Shengqu Cai, Anton Obukhov, Dengxin Dai, Luc van Gool

We propose a pipeline to generate Neural Radiance Fields (NeRF) of an object or a scene of a specific class, conditioned on a single input image.

3D-Aware Image Synthesis Novel View Synthesis +2

CoSSL: Co-Learning of Representation and Classifier for Imbalanced Semi-Supervised Learning

1 code implementation CVPR 2022 Yue Fan, Dengxin Dai, Anna Kukleva, Bernt Schiele

In this paper, we propose a novel co-learning framework (CoSSL) with decoupled representation learning and classifier learning for imbalanced SSL.

Representation Learning

Both Style and Fog Matter: Cumulative Domain Adaptation for Semantic Foggy Scene Understanding

no code implementations CVPR 2022 Xianzheng Ma, Zhixiang Wang, Yacheng Zhan, Yinqiang Zheng, Zheng Wang, Dengxin Dai, Chia-Wen Lin

Unlike previous methods that mainly focus on closing the domain gap caused by fog -- defogging the foggy images or fogging the clear images, we propose to alleviate the domain gap by considering fog influence and style variation simultaneously.

Disentanglement Domain Adaptation +1

DAFormer: Improving Network Architectures and Training Strategies for Domain-Adaptive Semantic Segmentation

3 code implementations CVPR 2022 Lukas Hoyer, Dengxin Dai, Luc van Gool

It improves the state of the art by 10. 8 mIoU for GTA-to-Cityscapes and 5. 4 mIoU for Synthia-to-Cityscapes and enables learning even difficult classes such as train, bus, and truck well.

Semantic Segmentation Synthetic-to-Real Translation +1

TACS: Taxonomy Adaptive Cross-Domain Semantic Segmentation

1 code implementation10 Sep 2021 Rui Gong, Martin Danelljan, Dengxin Dai, Danda Pani Paudel, Ajad Chhatkuli, Fisher Yu, Luc van Gool

In many real-world settings, the target domain task requires a different taxonomy than the one imposed by the source domain.

Contrastive Learning Domain Adaptation +1

Improving Semi-Supervised and Domain-Adaptive Semantic Segmentation with Self-Supervised Depth Estimation

1 code implementation28 Aug 2021 Lukas Hoyer, Dengxin Dai, Qin Wang, Yuhua Chen, Luc van Gool

Training deep networks for semantic segmentation requires large amounts of labeled training data, which presents a major challenge in practice, as labeling segmentation masks is a highly labor-intensive process.

Data Augmentation Domain Adaptation +5

End-to-End Urban Driving by Imitating a Reinforcement Learning Coach

2 code implementations ICCV 2021 Zhejun Zhang, Alexander Liniger, Dengxin Dai, Fisher Yu, Luc van Gool

Our end-to-end agent achieves a 78% success rate while generalizing to a new town and new weather on the NoCrash-dense benchmark and state-of-the-art performance on the challenging public routes of the CARLA LeaderBoard.

Autonomous Driving Imitation Learning +2

Fog Simulation on Real LiDAR Point Clouds for 3D Object Detection in Adverse Weather

1 code implementation ICCV 2021 Martin Hahner, Christos Sakaridis, Dengxin Dai, Luc van Gool

2) Through extensive experiments with several state-of-the-art detection approaches, we show that our fog simulation can be leveraged to significantly improve the performance for 3D object detection in the presence of fog.

3D Object Detection Object +3

Self-Aligned Video Deraining With Transmission-Depth Consistency

1 code implementation CVPR 2021 Wending Yan, Robby T. Tan, Wenhan Yang, Dengxin Dai

In this paper, we address the problems of rain streaks and rain accumulation removal in video, by developing a self-aligned network with transmission-depth consistency.

Optical Flow Estimation Rain Removal

ACDC: The Adverse Conditions Dataset with Correspondences for Semantic Driving Scene Understanding

no code implementations ICCV 2021 Christos Sakaridis, Dengxin Dai, Luc van Gool

To address this, we introduce ACDC, the Adverse Conditions Dataset with Correspondences for training and testing semantic segmentation methods on adverse visual conditions.

Scene Understanding Segmentation +2

Learnable Online Graph Representations for 3D Multi-Object Tracking

no code implementations23 Apr 2021 Jan-Nico Zaech, Dengxin Dai, Alexander Liniger, Martin Danelljan, Luc van Gool

Tracking of objects in 3D is a fundamental task in computer vision that finds use in a wide range of applications such as autonomous driving, robotics or augmented reality.

3D Multi-Object Tracking Autonomous Driving

Hyperspectral Image Super-Resolution with Spectral Mixup and Heterogeneous Datasets

2 code implementations19 Jan 2021 Ke Li, Dengxin Dai, Ender Konukoglu, Luc van Gool

With these contributions, our method is able to learn from heterogeneous datasets and lift the requirement for having a large amount of HD HSI training samples.

Data Augmentation Hyperspectral Image Super-Resolution +2

Three Ways to Improve Semantic Segmentation with Self-Supervised Depth Estimation

1 code implementation CVPR 2021 Lukas Hoyer, Dengxin Dai, Yuhua Chen, Adrian Köring, Suman Saha, Luc van Gool

Training deep networks for semantic segmentation requires large amounts of labeled training data, which presents a major challenge in practice, as labeling segmentation masks is a highly labor-intensive process.

Data Augmentation Monocular Depth Estimation +2

Cluster, Split, Fuse, and Update: Meta-Learning for Open Compound Domain Adaptive Semantic Segmentation

no code implementations CVPR 2021 Rui Gong, Yuhua Chen, Danda Pani Paudel, Yawei Li, Ajad Chhatkuli, Wen Li, Dengxin Dai, Luc van Gool

Open compound domain adaptation (OCDA) is a domain adaptation setting, where target domain is modeled as a compound of multiple unknown homogeneous domains, which brings the advantage of improved generalization to unseen domains.

Domain Adaptation Meta-Learning +2

Depth Estimation from Monocular Images and Sparse Radar Data

1 code implementation30 Sep 2020 Juan-Ting Lin, Dengxin Dai, Luc van Gool

We give a comprehensive study of the fusion between RGB images and Radar measurements from different aspects and proposed a working solution based on the observations.

Depth Estimation

Improving Point Cloud Semantic Segmentation by Learning 3D Object Detection

no code implementations22 Sep 2020 Ozan Unal, Luc van Gool, Dengxin Dai

Point cloud semantic segmentation plays an essential role in autonomous driving, providing vital information about drivable surfaces and nearby objects that can aid higher level tasks such as path planning and collision avoidance.

3D Object Detection 3D Semantic Segmentation +5

Multi-scale Interaction for Real-time LiDAR Data Segmentation on an Embedded Platform

2 code implementations20 Aug 2020 Shijie Li, Xieyuanli Chen, Yun Liu, Dengxin Dai, Cyrill Stachniss, Juergen Gall

Real-time semantic segmentation of LiDAR data is crucial for autonomously driving vehicles, which are usually equipped with an embedded platform and have limited computational resources.

Autonomous Vehicles Real-Time 3D Semantic Segmentation +1

Weakly Supervised 3D Object Detection from Lidar Point Cloud

1 code implementation ECCV 2020 Qinghao Meng, Wenguan Wang, Tianfei Zhou, Jianbing Shen, Luc van Gool, Dengxin Dai

This work proposes a weakly supervised approach for 3D object detection, only requiring a small set of weakly annotated scenes, associated with a few precisely labeled object instances.

3D Object Detection Object +1

Off-Policy Reinforcement Learning for Efficient and Effective GAN Architecture Search

1 code implementation ECCV 2020 Yuan Tian, Qin Wang, Zhiwu Huang, Wen Li, Dengxin Dai, Minghao Yang, Jun Wang, Olga Fink

In this paper, we introduce a new reinforcement learning (RL) based neural architecture search (NAS) methodology for effective and efficient generative adversarial network (GAN) architecture search.

Generative Adversarial Network Image Generation +3

Learning Accurate and Human-Like Driving using Semantic Maps and Attention

no code implementations10 Jul 2020 Simon Hecker, Dengxin Dai, Alexander Liniger, Luc van Gool

This paper investigates how end-to-end driving models can be improved to drive more accurately and human-like.

Analogical Image Translation for Fog Generation

no code implementations28 Jun 2020 Rui Gong, Dengxin Dai, Yu-Hua Chen, Wen Li, Luc van Gool

AIT achieves this zero-shot image translation capability by coupling a supervised training scheme in the synthetic domain, a cycle consistency strategy in the real domain, an adversarial training scheme between the two domains, and a novel network design.

Image-to-Image Translation Scene Understanding +1

Map-Guided Curriculum Domain Adaptation and Uncertainty-Aware Evaluation for Semantic Nighttime Image Segmentation

1 code implementation28 May 2020 Christos Sakaridis, Dengxin Dai, Luc van Gool

Our central contributions are: 1) a curriculum framework to gradually adapt semantic segmentation models from day to night through progressively darker times of day, exploiting cross-time-of-day correspondences between daytime images from a reference map and dark images to guide the label inference in the dark domains; 2) a novel uncertainty-aware annotation and evaluation framework and metric for semantic segmentation, including image regions beyond human recognition capability in the evaluation in a principled fashion; 3) the Dark Zurich dataset, comprising 2416 unlabeled nighttime and 2920 unlabeled twilight images with correspondences to their daytime counterparts plus a set of 201 nighttime images with fine pixel-level annotations created with our protocol, which serves as a first benchmark for our novel evaluation.

Domain Adaptation Image Segmentation +2

Action Sequence Predictions of Vehicles in Urban Environments using Map and Social Context

no code implementations29 Apr 2020 Jan-Nico Zaech, Dengxin Dai, Alexander Liniger, Luc van Gool

Our second contribution lies in applying the method to the well-known traffic agent tracking and prediction dataset Argoverse, resulting in 228, 000 action sequences.

Multi-Task Learning for Dense Prediction Tasks: A Survey

1 code implementation28 Apr 2020 Simon Vandenhende, Stamatios Georgoulis, Wouter Van Gansbeke, Marc Proesmans, Dengxin Dai, Luc van Gool

In this survey, we provide a well-rounded view on state-of-the-art deep learning approaches for MTL in computer vision, explicitly emphasizing on dense prediction tasks.

Multi-Task Learning

Quantifying Data Augmentation for LiDAR based 3D Object Detection

no code implementations3 Apr 2020 Martin Hahner, Dengxin Dai, Alexander Liniger, Luc van Gool

In this work, we shed light on different data augmentation techniques commonly used in Light Detection and Ranging (LiDAR) based 3D Object Detection.

3D Object Detection Data Augmentation +3

Semantic Object Prediction and Spatial Sound Super-Resolution with Binaural Sounds

no code implementations ECCV 2020 Arun Balajee Vasudevan, Dengxin Dai, Luc van Gool

We also propose two auxiliary tasks namely, a) a novel task on Spatial Sound Super-resolution to increase the spatial resolution of sounds, and b) dense depth prediction of the scene.

Depth Estimation Depth Prediction +1

Matching Neuromorphic Events and Color Images via Adversarial Learning

no code implementations2 Mar 2020 Fang Xu, ShiJie Lin, Wen Yang, Lei Yu, Dengxin Dai, Gui-Song Xia

The event camera has appealing properties: high dynamic range, low latency, low power consumption and low memory usage, and thus provides complementariness to conventional frame-based cameras.

Image Retrieval Retrieval

Don't Forget The Past: Recurrent Depth Estimation from Monocular Video

no code implementations8 Jan 2020 Vaishakh Patil, Wouter Van Gansbeke, Dengxin Dai, Luc van Gool

In particular, we put three different types of depth estimation (supervised depth prediction, self-supervised depth prediction, and self-supervised depth completion) into a common framework.

Depth Completion Depth Prediction +3

Self-supervised Object Motion and Depth Estimation from Video

no code implementations9 Dec 2019 Qi Dai, Vaishakh Patil, Simon Hecker, Dengxin Dai, Luc van Gool, Konrad Schindler

We present a self-supervised learning framework to estimate the individual object motion and monocular depth from video.

Depth Estimation Instance Segmentation +5

Texture Underfitting for Domain Adaptation

no code implementations29 Aug 2019 Jan-Nico Zaech, Dengxin Dai, Martin Hahner, Luc van Gool

Comprehensive semantic segmentation is one of the key components for robust scene understanding and a requirement to enable autonomous driving.

Autonomous Driving Domain Adaptation +4

Learning a Curve Guardian for Motorcycles

no code implementations12 Jul 2019 Simon Hecker, Alexander Liniger, Henrik Maurenbrecher, Dengxin Dai, Luc van Gool

Our contributes are fourfold: 1) we predict the motorcycle's intra-lane position using a convolutional neural network (CNN), 2) we predict the motorcycle roll angle using a CNN, 3) we use an upgraded controller model that incorporates road incline for a more realistic model and prediction, 4) we design a scale-able system by utilizing HERE Technologies map database to obtain the accurate road geometry of the future path.

Position

Gated CRF Loss for Weakly Supervised Semantic Image Segmentation

no code implementations11 Jun 2019 Anton Obukhov, Stamatios Georgoulis, Dengxin Dai, Luc van Gool

State-of-the-art approaches for semantic segmentation rely on deep convolutional neural networks trained on fully annotated datasets, that have been shown to be notoriously expensive to collect, both in terms of time and money.

Image Segmentation Weakly supervised Semantic Segmentation +1

Learning Accurate, Comfortable and Human-like Driving

no code implementations26 Mar 2019 Simon Hecker, Dengxin Dai, Luc van Gool

Our model is trained and evaluated on the Drive360 dataset, which features 60 hours and 3000 km of real-world driving data.

Autonomous Vehicles Navigate

Real-time 3D Traffic Cone Detection for Autonomous Driving

2 code implementations6 Feb 2019 Ankit Dhall, Dengxin Dai, Luc van Gool

In this work, we leverage the unique structure of traffic cones and propose a pipelined approach to the problem.

3D Object Detection Autonomous Driving +3

Guided Curriculum Model Adaptation and Uncertainty-Aware Evaluation for Semantic Nighttime Image Segmentation

1 code implementation ICCV 2019 Christos Sakaridis, Dengxin Dai, Luc van Gool

Our central contributions are: 1) a curriculum framework to gradually adapt semantic segmentation models from day to night via labeled synthetic images and unlabeled real images, both for progressively darker times of day, which exploits cross-time-of-day correspondences for the real images to guide the inference of their labels; 2) a novel uncertainty-aware annotation and evaluation framework and metric for semantic segmentation, designed for adverse conditions and including image regions beyond human recognition capability in the evaluation in a principled fashion; 3) the Dark Zurich dataset, which comprises 2416 unlabeled nighttime and 2920 unlabeled twilight images with correspondences to their daytime counterparts plus a set of 151 nighttime images with fine pixel-level annotations created with our protocol, which serves as a first benchmark to perform our novel evaluation.

Image Segmentation Segmentation +2

Curriculum Model Adaptation with Synthetic and Real Data for Semantic Foggy Scene Understanding

1 code implementation5 Jan 2019 Dengxin Dai, Christos Sakaridis, Simon Hecker, Luc van Gool

The method is based on the fact that the results of semantic segmentation in moderately adverse conditions (light fog) can be bootstrapped to solve the same problem in highly adverse conditions (dense fog).

Domain Adaptation Scene Understanding +2

Model Adaptation with Synthetic and Real Data for Semantic Dense Foggy Scene Understanding

no code implementations ECCV 2018 Christos Sakaridis, Dengxin Dai, Simon Hecker, Luc van Gool

In addition, we present three other main stand-alone contributions: 1) a novel method to add synthetic fog to real, clear-weather scenes using semantic input; 2) a new fog density estimator; 3) the Foggy Zurich dataset comprising $3808$ real foggy images, with pixel-level semantic annotations for $16$ images with dense fog.

Scene Understanding Semantic Segmentation

Unified Hypersphere Embedding for Speaker Recognition

1 code implementation22 Jul 2018 Mahdi Hajibabaei, Dengxin Dai

Incremental improvements in accuracy of Convolutional Neural Networks are usually achieved through use of deeper and more complex models trained on larger datasets.

Speaker Recognition Text-Independent Speaker Recognition

Failure Prediction for Autonomous Driving

no code implementations4 May 2018 Simon Hecker, Dengxin Dai, Luc van Gool

This work presents a method to learn to predict the occurrence of these failures, i. e. to assess how difficult a scene is to a given driving model and to possibly give the human driver an early headsup.

Autonomous Driving

End-to-End Learning of Driving Models with Surround-View Cameras and Route Planners

no code implementations ECCV 2018 Simon Hecker, Dengxin Dai, Luc van Gool

In particular, we develop a sensor setup that provides data for a 360-degree view of the area surrounding the vehicle, the driving route to the destination, and low-level driving maneuvers (e. g. steering angle and speed) by human drivers.

Learning the Synthesizability of Dynamic Texture Samples

no code implementations3 Feb 2018 Feng Yang, Gui-Song Xia, Dengxin Dai, Liangpei Zhang

In this paper, we investigate the synthesizability of dynamic texture samples: {\em given a dynamic texture sample, how synthesizable it is by using EDTS, and which EDTS method is the most suitable to synthesize it?}

regression Texture Synthesis

Object Referring in Videos with Language and Human Gaze

no code implementations CVPR 2018 Arun Balajee Vasudevan, Dengxin Dai, Luc van Gool

To that end, we present a new video dataset for OR, with 30, 000 objects over 5, 000 stereo video sequences annotated for their descriptions and gaze.

Object Referring Expression

Object Referring in Visual Scene with Spoken Language

no code implementations10 Nov 2017 Arun Balajee Vasudevan, Dengxin Dai, Luc van Gool

This paper investigates Object Referring with Spoken Language (ORSpoken) by presenting two datasets and one novel approach.

Object

Semantic Foggy Scene Understanding with Synthetic Data

no code implementations25 Aug 2017 Christos Sakaridis, Dengxin Dai, Luc van Gool

Due to the difficulty of collecting and annotating foggy images, we choose to generate synthetic fog on real images that depict clear-weather outdoor scenes, and then leverage these partially synthetic data for SFSU by employing state-of-the-art convolutional neural networks (CNN).

Image Dehazing object-detection +3

Deep Domain Adaptation by Geodesic Distance Minimization

no code implementations13 Jul 2017 Yifei Wang, Wen Li, Dengxin Dai, Luc van Gool

Our work builds on the recently proposed Deep CORAL method, which proposed to train a convolutional neural network and simultaneously minimize the Euclidean distance of convariance matrices between the source and target domains.

Domain Adaptation

Speech-Based Visual Question Answering

1 code implementation1 May 2017 Ted Zhang, Dengxin Dai, Tinne Tuytelaars, Marie-Francine Moens, Luc van Gool

This paper introduces speech-based visual question answering (VQA), the task of generating an answer given an image and a spoken question.

Automatic Speech Recognition Automatic Speech Recognition (ASR) +3

PathTrack: Fast Trajectory Annotation with Path Supervision

no code implementations ICCV 2017 Santiago Manen, Michael Gygli, Dengxin Dai, Luc van Gool

We further validate our approach by crowdsourcing the PathTrack dataset, with more than 15, 000 person trajectories in 720 sequences.

Multiple Object Tracking Object +1

Fast Algorithms for Linear and Kernel SVM+

no code implementations CVPR 2016 Wen Li, Dengxin Dai, Mingkui Tan, Dong Xu, Luc van Gool

The SVM+ approach has shown excellent performance in visual recognition tasks for exploiting privileged information in the training data.

Scale-Aware Alignment of Hierarchical Image Segmentation

1 code implementation CVPR 2016 Yuhua Chen, Dengxin Dai, Jordi Pont-Tuset, Luc van Gool

To demonstrate the power of our method, we perform comprehensive experiments, which show that our method, as a post-processing step, can significantly improve the quality of the hierarchical segmentation representations, and ease the usage of hierarchical image segmentation to high-level vision tasks such as object segmentation.

Image Segmentation Segmentation +1

Fast Optical Flow using Dense Inverse Search

no code implementations11 Mar 2016 Till Kroeger, Radu Timofte, Dengxin Dai, Luc van Gool

Most recent works in optical flow extraction focus on the accuracy and neglect the time complexity.

Action Detection Activity Detection +1

Unsupervised High-level Feature Learning by Ensemble Projection for Semi-supervised Image Classification and Image Clustering

no code implementations2 Feb 2016 Dengxin Dai, Luc van Gool

Hence, in the spirit of ensemble learning we create a set of such training sets which are all diverse, leading to diverse classifiers.

Classification Clustering +6

Is Image Super-resolution Helpful for Other Vision Tasks?

no code implementations23 Sep 2015 Dengxin Dai, Yujian Wang, Yuhua Chen, Luc van Gool

In this paper, we present the first comprehensive study and analysis of the usefulness of ISR for other vision applications.

Edge Detection Image Segmentation +4

Joint Vanishing Point Extraction and Tracking

no code implementations CVPR 2015 Till Kroeger, Dengxin Dai, Luc van Gool

Although the method is designed for unknown camera poses, it is also helpful in scenarios with known poses, since a multi-frame approach in VP detection helps to regularize in frames with weak VP line support.

Metric Imitation by Manifold Transfer for Efficient Vision Applications

no code implementations CVPR 2015 Dengxin Dai, Till Kroeger, Radu Timofte, Luc van Gool

In particular, MI consists of: 1) quantifying the properties of source metrics as manifold geometry, 2) transferring the manifold from source domain to target domain, and 3) learning a mapping of TFs so that the manifold is approximated as well as possible in the mapped feature domain.

Clustering Image Clustering +5

The Synthesizability of Texture Examples

no code implementations CVPR 2014 Dengxin Dai, Hayko Riemenschneider, Luc van Gool

This work is the first attempt to quantify this image property, and we find that texture synthesizability can be learned and predicted.

Texture Synthesis

Latent Dictionary Learning for Sparse Representation based Classification

no code implementations CVPR 2014 Meng Yang, Dengxin Dai, Lilin Shen, Luc van Gool

Each dictionary atom is jointly learned with a latent vector, which associates this atom to the representation of different classes.

Classification Dictionary Learning +4

Cannot find the paper you are looking for? You can Submit a new open access paper.