Predicting performance of object recognition

Boshra, M.; Bhanu, Bir

doi:10.1109/34.877519

Cited by 40 publications

(28 citation statements)

References 25 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This Risk-Averse Metric (RAM) gives BASESYS +1.0 points for a correct answer, -0.5 points for an incorrect answer and 0 points if it makes no decision. 4 We believe such a metric is crucial when dealing with real applications. By refusing to make a decision when ALERT raises a warning, we expect BASESYS to gain more points by trading off incorrect decisions for no decisions.…”

Section: Discussionmentioning

confidence: 99%

“…Hoiem et al [24] focus on analyzing the different sources of error in object detectors, and do not predict failure. Methods that predict performance by analyzing statistics of the training and test data [4,57] are not applicable to our goal of predicting the reliability of individual test instances. Detecting errors has received a lot of attention in speech recognition [8,45].…”

Section: Related Workmentioning

confidence: 99%

See 1 more Smart Citation

Predicting Failures of Vision Systems

Zhang

Wang

Farhadi

et al. 2014

2014 IEEE Conference on Computer Vision and Pattern Recognition

117

View full text Add to dashboard Cite

Computer vision systems today fail frequently. They also fail abruptly without warning or explanation. Alleviating the former has been the primary focus of the community. In this work, we hope to draw the community's attention to the latter, which is arguably equally problematic for real applications. We promote two metrics to evaluate failure prediction. We show that a surprisingly straightforward and general approach, that we call ALERT, can predict the likely accuracy (or failure) of a variety of computer vision systems -semantic segmentation, vanishing point and camera parameter estimation, and image memorability predictionon individual input images. We also explore attribute prediction, where classifiers are typically meant to generalize to new unseen categories. We show that ALERT can be useful in predicting failures of this transfer. Finally, we leverage ALERT to improve the performance of a downstream application of attribute prediction: zero-shot learning. We show that ALERT can outperform several strong baselines for zero-shot learning on four datasets.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

Predicting Failures of Vision Systems

Zhang

Wang

Farhadi

et al. 2014

2014 IEEE Conference on Computer Vision and Pattern Recognition

117

View full text Add to dashboard Cite

show abstract

“…Global matching (e.g., silhouette image matching) approaches consider finding a transformation from a model to an image while feature matching approaches involve establishing a correspondence between local features extracted from the given data and corresponding local features of the object model. Boshra and Bhanu (2000) present a method for predicting fundamental performance of object recognition. They assume that both scene data and model objects are represented by 2D point features and a data/model match is evaluated using a votebased criterion.…”

Section: Introductionmentioning

confidence: 99%

Performance prediction for individual recognition by gait

Han

Bhanu

2005

Pattern Recognition Letters

View full text Add to dashboard Cite

“…In model-based object recognition, (Jones & Bhanu, 1999) described a model-based object recognition method using the combination of a SAR approach, model for azimuthal variance, articulation invariants, and the resolution of the sensor data. (Boshra & Bhanu, 2000) also described a model-based object recognition method using the probability of correct recognition. Current methods for dealing with occlusion have been based on template matching, statistical approaches using localized invariants, and recognition of occluded regions based on local features.…”

Section: Related Workmentioning

confidence: 99%