A Benchmark of Computational Models of Saliency to Predict Human Fixations

By Tilke Judd, Frédo Durand, Antonio Torralba. 2012

630 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition, Evaluation & Benchmarks

How researchers see this paper

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

In Computer Vision and Pattern Recognition

In Evaluation & Benchmarks

Citations in the corpus

Make-A-Scene: Scene-Based Text-to-Image Generation with Human Priors (2022, background): “A clear example of this gap can be observed in person and face generation, where a dissonance is present between the importance of face pixels from the human perspective and the loss applied over the whole image [28, 66].”
Faster gaze prediction with dense networks and Fisher pruning (2018, methodology): “To verify that our models indeed perform close to the state of the art, we submitted saliency maps to the MIT Saliency Benchmark [3, 14, 4].”
Visual saliency estimation by nonlinearly integrating features using region covariances. (2013, methodology): “Earth Movers Distance (EMD) (Pele & Werman, 2009; Rubner, Tomasi, & Guibas, 2000), and (d) the similarity score suggested by Judd et al. (2012). In the next subsection, we report the mean values of these metrics averaged over all the images in the data sets. The ROC score is the most commonly used m”
AGIL: Learning Attention from Human for Visuomotor Tasks (2018, background): “In recent years, this approach has made tremendous progress due to large benchmark datasets such as [20, 18] and deep neural networks [4, 24, 21, 41].”
Saliency and Human Fixations: State-of-the-Art and Study of Comparison Metrics (2013, methodology): “[14] used this metric in her benchmark which is now available online.”
Learning Uncertain Convolutional Features for Accurate Saliency Detection (2017, methodology): “AUC-J sAUC CC NSS IG MIT300 [19] 0.8584 0.7109 0.7423 2.14 - iSUN [47] 0.8615 0.5498 0.8142 - 0.1725 SALICON[17] 0.7621 0.6326 0.8453 - 0.3167 Table 3.”
Global Contrast Based Salient Region Detection (2015, background): “In contrast, computationally identifying such salient object regions [2], [3], that match the human annotators’ behaviour when they have been asked to pick a salient object in an image, is very challenging.”
Efficient Salient Region Detection with Soft Image Abstraction (2013, methodology): “As correctly pointed out by recent survey papers, there are two major branches of visual attention modeling: saliency region detection [7] and human fixation prediction [6, 32].”
A deep multi-level network for saliency prediction (2016, methodology): “The Similarity metric [28] computes the sum of pixel-wise minimums between the predicted saliency map SM and the human eye ﬁxation map FM , where SM and FM are supposed to be probability distributions and sum up to one.”
Deep Gaze I: Boosting Saliency Prediction with Feature Maps Trained on ImageNet (2014, methodology): “We submitted our model to the MIT Saliency Benchmark (Judd et al., 2012).”
Recurrent Mixture Density Network for Spatiotemporal Visual Attention (2016, methodology): “We evaluate all models on the test set, using popular metrics proposed in the literature of saliency map prediction for still images (Judd et al., 2012; Borji et al., 2013), such as Area Under the ROC Curve (AUC), Normalized Scanpath Saliency (NSS), linear Correlation Coefficient (CC) and the Simila”
End-to-end Convolutional Network for Saliency Prediction (2015, methodology): “The evaluation metrics was adopted of the variety of metrics provided in MIT saliency benchmark [13, 4] defined on both saliency map and fixation points.”
Salient Object Detection: A Discriminative Regional Feature Integration Approach (2013, background): “Eye fixation prediction, another visual saliency research direction, also attracts a lot of interests [7, 24].”
Visual Saliency Based on Scale-Space Analysis in the Frequency Domain (2013, methodology): “Note that is not the same as the center-bias (CB) or border cut (BC) described in [28, since it is used only to choose a proper scale, but not to modify the saliency map as done in [7], [44].”
Large-Scale Optimization of Hierarchical Features for Saliency Prediction in Natural Images (2014, methodology): “ter match the distribution of eye movements [17].”
DeepFix: A Fully Convolutional Neural Network for Predicting Human Eye Fixations (2015, methodology): “MIT300 : This dataset [5] contains 300 natural images from both indoor and outdoor scenarios.”