Recent research has shown that faces can be obfuscated in large-scale datasets with a minimal performance impact on image classification and downstream tasks like object recognition. In this paper, we investigate the ...
详细信息
ISBN:
(纸本)9781665448994
Recent research has shown that faces can be obfuscated in large-scale datasets with a minimal performance impact on image classification and downstream tasks like object recognition. In this paper, we investigate the role of face obfuscation in video classification datasets and quantify a more significant reduction in performance caused by face blurring. To reduce such performance effects, we propose a generalized distillation approach in which a privacy-preserving action recognition network is trained with privileged information given by face identities. We show, through experiments performed on Kinetics-400, that the proposed approach can fully close the performance gap caused by face anonymization.
We present an approach to perform supervised action recognition in the dark. In this work, we present our results on the ARID dataset[60]. Most previous works only evaluate performance on large, well illuminated datas...
详细信息
ISBN:
(纸本)9781665448994
We present an approach to perform supervised action recognition in the dark. In this work, we present our results on the ARID dataset[60]. Most previous works only evaluate performance on large, well illuminated datasets like Kinetics and HMDB51. We demonstrate that our work is able to achieve a very low error rate while being trained on a much smaller dataset of dark videos. We also explore a variety of training and inference strategies including domain transfer methodologies and also propose a simple but useful frame selection strategy. Our empirical results demonstrate that we beat previously published baseline models by 11%.
This paper introduces a novel dataset for video enhancement and studies the state-of-the-art methods of the NTIRE 2021 challenge on quality enhancement of compressed video. The challenge is the first NTIRE challenge i...
详细信息
ISBN:
(纸本)9781665448994
This paper introduces a novel dataset for video enhancement and studies the state-of-the-art methods of the NTIRE 2021 challenge on quality enhancement of compressed video. The challenge is the first NTIRE challenge in this direction, with three competitions, hundreds of participants and tens of proposed solutions. Our newly collected Large-scale Diverse Video (LDV) dataset is employed in the challenge. In our study, we analyze the solutions of the challenges and several representative methods from previous literature on the proposed LDV dataset. We find that the NTIRE 2021 challenge advances the state-of-theart of quality enhancement on compressed video.
Several vision problems can be reduced to the problem of fitting a linear surface of low dimension to data, including the problems of structure-from-affine-motion, and of characterizing the intensity images of a Lambe...
详细信息
ISBN:
(纸本)0780342364
Several vision problems can be reduced to the problem of fitting a linear surface of low dimension to data, including the problems of structure-from-affine-motion, and of characterizing the intensity images of a Lambertian scene by constructing the intensity manifold. For these problems, one must deal with a data matrix with some missing elements. In structure-from-motion, missing elements will occur if some point features are not visible in some frames. To construct the intensity manifold missing matrix elements will arise when the surface normals of some scene points do not face the light source in some images. We propose a novel method for fitting a low rank matrix to a matrix with missing elements. We show experimentally that our method produces good results in the presence of noise. These results can be either used directly, or can serve as an excellent starting point for an iterative method.
During the performance optimization of a computervision system, developers frequently run into platform-level inefficiencies and bottlenecks that can not be addressed by traditional methods. OpenVX is designed to add...
详细信息
ISBN:
(纸本)9781479943098
During the performance optimization of a computervision system, developers frequently run into platform-level inefficiencies and bottlenecks that can not be addressed by traditional methods. OpenVX is designed to address such system-level issues by means of a graph-based computation model. This approach differs from the traditional acceleration of one-off functions, and exposes optimization possibilities that might not be available or obvious with traditional computervision libraries such as OpenCV.
The choice of a color space is of great importance for many computervision algorithms (e.g. edge detection and object recognition). It induces the equivalence classes to the actual algorithms. However the problem is ...
详细信息
ISBN:
(纸本)0769523722
The choice of a color space is of great importance for many computervision algorithms (e.g. edge detection and object recognition). It induces the equivalence classes to the actual algorithms. However the problem is how to automatically select the color space that produces the best result for a particular task. The subsequent difficulty then is how to obtain a proper weighting scheme for the algorithms so that the results are combined in an optimal setting. To achieve proper color space selection and fusion of feature detectors, in this paper we propose a method that exploits non-perfect correlation between the color models derived from the principles of diversification. As a consequence, the weighting scheme yields maximal color discrimination. The method is verified experimentally for two different feature detectors. The experimental results show that the model provides feature detection results having a discriminative power of 30 percent higher than the standard weighting scheme.
Towards the goal of realizing a generic automatic human activity recognition system, a new formalism is proposed. Activities are described by a chained hierarchical representation using three type of entities: image f...
详细信息
ISBN:
(纸本)0769506623
Towards the goal of realizing a generic automatic human activity recognition system, a new formalism is proposed. Activities are described by a chained hierarchical representation using three type of entities: image features, mobile object properties and scenarios. Taking image features of tracked moving regions from an image sequence as input, mobile object properties are first computed by specific methods ods while noise is suppressed by statistical methods. Scenarios are recognized from mobile object properties based on Bayesian analysis. A sequential occurance several scenarios are recognized by an algorithm using a probabilistic finite-state automation (a variant of structured HMM). The demonstration of the optimality of these recognition method is discussed. Finally, the validity and the effectiveness of our approach is demonstrated on both real-world and perturbed data.
We show how to outsource data annotation to Amazon Mechanical Turk. Doing so has produced annotations in quite large numbers relatively cheaply. The quality is good, and can be checked and controlled. Annotations are ...
详细信息
ISBN:
(纸本)9781424423392
We show how to outsource data annotation to Amazon Mechanical Turk. Doing so has produced annotations in quite large numbers relatively cheaply. The quality is good, and can be checked and controlled. Annotations are produced quickly. We describe results for several different annotation problems. We describe some strategies for determining when the task is well specified and properly priced.
Recognizing a target in synthetic-aperture radar (SAR) images is an important, yet challenging, application of the model-based vision technique. This paper describes a model-based SAR recognition system based on invar...
详细信息
ISBN:
(纸本)0818672587
Recognizing a target in synthetic-aperture radar (SAR) images is an important, yet challenging, application of the model-based vision technique. This paper describes a model-based SAR recognition system based on invariant histograms and deformable template matching techniques. An invariant histogram is a histogram of invariant values defined by geometric features such as points and lines in SAR images. Although a few invariants are sufficient to recognize a target, we use a histogram of all invariant values given by all possible target feature pairs. This redundant histogram enables robust recognition under severe occlusions typical in SAR recognition scenarios. Multi-step deformable template matching examines the existence of an object by superimposing templates over potential energy field generated from images or primitive features. It determines the template configuration which has the minimum deformation and the best alignment of the template with features. The deformability of the template absorbs the instability of SAR features. We have implemented the system and evaluated the system performance using hybrid SAR images, generated from synthesized model signatures and real SAR background signatures.
Image anonymization is widely adapted in practice to comply with privacy regulations in many regions. However, anonymization often degrades the quality of the data, reducing its utility for computervision development...
详细信息
ISBN:
(纸本)9798350302493
Image anonymization is widely adapted in practice to comply with privacy regulations in many regions. However, anonymization often degrades the quality of the data, reducing its utility for computervision development. In this paper, we investigate the impact of image anonymization for training computervision models on key computervision tasks (detection, instance segmentation, and pose estimation). Specifically, we benchmark the recognition drop on common detection datasets, where we evaluate both traditional and realistic anonymization for faces and full bodies. Our comprehensive experiments reflect that traditional image anonymization substantially impacts final model performance, particularly when anonymizing the full body. Furthermore, we find that realistic anonymization can mitigate this decrease in performance, where our experiments reflect a minimal performance drop for face anonymization. Our study demonstrates that realistic anonymization can enable privacy-preserving computervision development with minimal performance degradation across a range of important computervision benchmarks.
暂无评论