Computer Vision and Image Understanding

This paper presents a novel semi-supervised few-shot image classification method named Learning to Teach and Learn (LTTL) to effectively leverage unlabeled samples in small-data regimes. Our method is based on self-training, which ...

• We propose a novel self-training strategy for semi-supervised few-shot image classification.

In this work, we define and address a novel domain adaptation (DA) problem in semantic scene segmentation, where the target domain not only exhibits a data distribution shift w.r.t. the source domain, but also includes novel classes ...

• Proposing a new task in semantic segmentation where the target domain includes novel classes that do not exist in the source domain.

The correlation filter (CF) achieves excellent performance, showing high robustness to motion blur or illumination change by learning filters. However, tracking in challenging scenarios with occlusion or out-of-view is still not well ...

• The training samples of the CF framework are reconstructed within a subspace in our method to alleviate the problem of occlusion or out-of-view.

Semantic segmentation is one of the most fundamental problems in computer vision with significant impact on a wide variety of applications. Adversarial learning is shown to be an effective approach for improving semantic segmentation ...

• We propose lookahead adversarial learning (LoAd) for adversarial semantic segmentation.

There hardly exists any large-scale datasets with dense optical flow of non-rigid motion from real-world imagery as of today. The reason lies mainly in the required setup to derive ground truth optical flows: a series of images with ...

Display Omitted

• First method to automatically generate dense non-rigid optical flow data for training.

Human pose estimation is a very active research field, stimulated by its important applications in robotics, entertainment or health and sports sciences, among others. Advances in convolutional networks triggered noticeable ...

• Survey of recent years 3D pose estimation methods.
• Analysis based on ...

Numerous practical datasets have been developed to recognize human actions from videos. However, many of them were constructed by collecting videos within a limited domain; thus, a model trained using one of the existing datasets often ...

• We construct a meta video dataset for human action recognition, called MetaVD.
• ...

A precise large-scale dataset is crucial for supervising the training of deep neural networks (DNNs) in image classification. However, manually annotating large-scale dataset is time-consuming, which limits the scalability of ...

Display Omitted

• For reducing the impact of label noise, this work studies how to effectively and efficiently train deep networks on the noisy large-scale dataset in ...

Matrix regression uses matrix data as input and directly selects the features from matrix data by employing several couples of left and right regression matrices. However, the existing matrix regression methods do not consider the ...

• Our model preserves the spatial information of elements in the original image data.