From Virtual to Reality: Fast Adaptation of Virtual Object Detectors to Real Domains

By Baochen Sun, Kate Saenko. 2014

169 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computer Vision and Pattern Recognition

Citations in the corpus

Unsupervised Domain Adaptation by Backpropagation (2014, background): “One particularly important example is synthetic or semi-synthetic imagery, which may come in abundance and fully labeled, but which inevitably look different from real data [13, 20, 23, 21].”
Recognition in Terra Incognita (2018, methodology): “[48,49,50,51,52] look at methods for adapting detectors from one data source to another, such as from synthetic to real data or from images to video.”
Single Image 3D Interpreter Network (2016, methodology): “[39] used the same strategy for object detection.”
Deep CORAL: Correlation Alignment for Deep Domain Adaptation (2016, background): “Instead of collecting labelled data and training a new classiﬁer for every possible scenario, unsupervised domain adaptation methods [6,7,8,9,10,1] try to compensate for the degradation in performance by transferring knowledge from labelled source domains to unlabelled target domains.”
Domain Adaptation for Visual Applications: A Comprehensive Survey (2017, methodology): “mount of annotated data. Synthetic data based adaptation. Early methods use 3D CAD models to improve solutions for pose and viewpoint estimation [161, 162, 163, 164], object and object part detection [165, 166, 167, 168, 169, 170, 171, 172], segmentation and scene understanding [173, 174, 175]. The ”
Return of Frustratingly Easy Domain Adaptation (2015, methodology): “In Tables 3 and 4, we compare CORAL to SA, GFK, TCA which have available source code as well as the NA baseline.”
Adapting Deep Visuomotor Representations with Weak Pairwise Constraints (2015, methodology): “n the two distributions [28], [17]. Recently, a method has been proposed to use 3D object models to render synthetic training examples for training visual models with limited human annotations needed [29]. It was shown that there is a speciﬁc domain shift problem that arises when applying a syntheti”
Domain Adaptive Faster R-CNN for Object Detection in the Wild (2018, background): “There also exists work to learn detectors from alternative sources, such as from images to videos [54], from 3D models [46, 53], or from synthetic models [25].”
A Unified Perspective on Multi-Domain and Multi-Task Learning (2014, background): “A variety of studies have proposed both supervised (Saenko et al., 2010; Duan et al., 2012) and unsupervised (Gong et al., 2012; Sun & Saenko, 2014) methods.”
Synthetic Data for Deep Learning (2019, background): “Sun and Saenko [567] propose a domain adaptation approach based on decorrelating the features of a classifier, both”
Learning scene-specific pedestrian detectors without real data (2015, unknown): “Sun and Saenko [34] used virtual objects to train 2D object detector for real objects.”
A self-supervised learning system for object detection using physics simulation and multi-view pose estimation (2017, background): “lighting, and shape to address this disparity [8].”
Towards Adapting Deep Visuomotor Representations from Simulated to Real Environments (2015, methodology): “Recently, a method has been proposed to use 3D object models to render synthetic training examples for training visual models with limited human annotations needed [29].”
Exploring Invariances in Deep Convolutional Neural Networks Using Synthetic Images (2014, methodology): “Previous work by Sun & Saenko (2014) has shown that when training a HOG model from virtual data, rendering natural backgrounds and texture is not helpful.”
Co-regularized Alignment for Unsupervised Domain Adaptation (2018, background): “Domain adaptation finds applications in many practical scenarios, including the special case when source domain consists of simulated or synthetic data (for which labels are readily available from the simulator) and target domain consists of real world observations [43, 39, 5].”
Play and Learn: Using Video Games to Train Computer Vision Models (2016, methodology): “In the recent literature, a variety of methods [4, 22, 30, 43] use 3d CAD models with simple rendering to tackle vision problems.”