Cycle-Consistent Domain Adaptive Faster RCNN

Zhang, Dan; Li, Jingjing; Xiong, Lin; Lin, Lan; Ye, Mao; Yang, Shangming

doi:10.1109/access.2019.2938837

Cited by 29 publications

(15 citation statements)

References 42 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…3.1.2) by adding an image-to-image translation at the input level while adding other losses like gradient reversal at an instance and image-level intact. As shown in their experiments [110], the addition of an image-translation module further enhances the performance.…”

Section: Image-to-image Translationmentioning

confidence: 82%

“…Arruda et al [112] Zhang et al [110] Graph reasoning Xu et al [97] Zhao et al [95] Sovinay et al [104]…”

Section: Pseudo-label Self-trainingmentioning

confidence: 99%

“…Image-toimage translation helps in bridging the gap at the input level while feature level adversarial alignment ensures a shared feature space for the target and source domain. Zhang et al [110] specifically utilize Cycle-GAN to learn a mapping function between source and target domain images, in a method termed as Cycle-consistent Adaptive Faster-RCNN (CA-FRCNN) and illustrated in Fig. 18.…”

Section: Image-to-image Translationmentioning

confidence: 99%

See 2 more Smart Citations

Unsupervised Domain Adaptation of Object Detectors: A Survey

Oza¹,

Sindagi²,

VS³

et al. 2021

Preprint

View full text Add to dashboard Cite

Recent advances in deep learning have led to the development of accurate and efficient models for various computer vision applications such as object classification, semantic segmentation, and object detection. However, learning highly accurate models relies on the availability of datasets with large number of annotated images. Due to this, model performance drops drastically when evaluated on label-scarce datasets having visually distinct images. This issue is commonly referred to as covariate shift or dataset bias. Domain adaptation attempts to address this problem by leveraging domain shift characteristics from labeled data in a related domain when learning a classifier for label-scarce target dataset. There are a plethora of works to adapt object classification and semantic segmentation models to label-scarce target dataset through unsupervised domain adaptation. Considering that object detection is a fundamental task in computer vision, many recent works have recently focused on addressing the domain adaptation issue for object detection as well. In this paper, we provide a brief introduction to the domain adaptation problem for object detection and present an overview of various methods proposed till date for addressing this problem. Furthermore, we highlight strategies proposed for this problem and the associated shortcomings. Subsequently, we identify multiple aspects of the unsupervised domain adaptive detection problem that are most promising for the future research in the area. We believe that this survey shall be valuable to the pattern recognition experts working in the fields of computer vision, biometrics, medical imaging, and autonomous navigation by introducing them to the problem, getting them familiar with the current status of the progress and providing with promising direction for future research.

show abstract

Section: Image-to-image Translationmentioning

confidence: 82%

“…Arruda et al [112] Zhang et al [110] Graph reasoning Xu et al [97] Zhao et al [95] Sovinay et al [104]…”

Section: Pseudo-label Self-trainingmentioning

confidence: 99%

Section: Image-to-image Translationmentioning

confidence: 99%

See 1 more Smart Citation

Unsupervised Domain Adaptation of Object Detectors: A Survey

Oza¹,

Sindagi²,

VS³

et al. 2021

Preprint

View full text Add to dashboard Cite

show abstract

“…More recently, adversarial training is actively used to reduce the domain gap in the deep learning framework. Those works are applying adversarial training at image level and instance level [42], aligning features strongly at a low-level feature and weakly at a high-level feature [63], adding cycle-consistent constraint for preserving identity [64], and progressively reducing domain gap via intermediate domain at image level [65] or score level [62].…”

Section: Related Workmentioning

confidence: 99%

Cut-and-Paste Dataset Generation for Balancing Domain Gaps in Object Instance Detection

et al. 2021

View full text Add to dashboard Cite

Training an object instance detector where only a few training object images are available is a challenging task. One solution is a cut-and-paste method that generates a training dataset by cutting object areas out of training images and pasting them onto other background images. A detector trained on a dataset generated with a cut-and-paste method suffers from the conventional domain shift problem, which stems from a discrepancy between the source domain (generated training dataset) and the target domain (real test dataset). Though state-of-the-art domain adaptation methods are able to reduce this gap, it is limited because they do not consider the difference of domain gaps of foreground and background. In this study, we present that the conventional domain gap can be divided into two sub-domain gaps for foreground and background. Then, we show that the original cut-and-paste approach suffers from a new domain gap problem, an unbalanced domain gaps, because it has two separate source domains for foreground and background, unlike the conventional domain shift problem. Then, we introduce an advanced cut-and-paste method to balance the unbalanced domain gaps by diversifying the foreground with GAN (generative adversarial network)-generated seed images and simplifying the background using image processing techniques. Experimental results show that our method is effective for balancing domain gaps and improving the accuracy of object instance detection in a cluttered indoor environment using only a few seed images. Furthermore, we show that balancing domain gaps can improve the detection accuracy of state-of-the-art domain adaptation methods.INDEX TERMS Artificial neural networks, image processing, learning (artificial intelligence), object detection.

show abstract

“…Faster RCNN is one of the most representative methods in the object detection field [32]. Considering the limited computational resources of the humanoid robots in the experiments and the disadvantages of the object recognition in the original network, such as the low recognition rate of the small object detection [33,34], the original network is optimized to balance the computational cost and the detection accuracy.…”

Section: Object Detection and Simulationmentioning

confidence: 99%

Optimized Convolutional Neural Network-Based Object Recognition for Humanoid Robot

Xiao¹,

Zeng²,

Yuan³

2020

J Robotics Autom

View full text Add to dashboard Cite

In recent years, many researchers have proposed a series of algorithms based on convolutional neural networks and achieved good performances in the field of object detection and recognition. For humanoid robots, they are designed to assist or replace people in completing a series of anthropomorphic tasks, and their ability to recognize and grasp surrounding objects is the most basic requirement. Therefore, this paper proposes an algorithm based on the optimized convolutional neural network (CNN) and can be used to detect and classify the objects by humanoid robots. Particularly, Faster region-based CNN (Faster R-CNN) is used for detecting the objects and the original network is refined by feature concatenation strategy. The feature fusion layer is also proposed to make the feature map contain more information, the improved non-maximum suppression (NMS) algorithm is adopted to solve the overlapping object detection problem, and some layers are also modified in the original network to get faster speed and smaller network size. The recognition and grasping performance of the humanoid robots are tested by the proposed algorithms in a virtual simulation environment, and the experiment results show that the proposed algorithms are feasible and perform well.

show abstract

Cycle-Consistent Domain Adaptive Faster RCNN

Cited by 29 publications

References 42 publications

Unsupervised Domain Adaptation of Object Detectors: A Survey

Unsupervised Domain Adaptation of Object Detectors: A Survey

Cut-and-Paste Dataset Generation for Balancing Domain Gaps in Object Instance Detection

Optimized Convolutional Neural Network-Based Object Recognition for Humanoid Robot

Contact Info

Product

Resources

About