How to Evaluate Foreground Maps

By Ran Margolin, Lihi Zelnik-Manor, Ayellet Tal. 2014 IEEE Conference on Computer Vision and Pattern Recognition, 2014

1,036 citations recorded by the source corpus.

View Paper: https://www.semanticscholar.org/paper/eaaa16fe11bf17b6df58dfbd346a5b8cbfabd596

How researchers see this paper

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

Citations in the corpus

Supervised Evaluation of Image Segmentation and Object Proposal Techniques (2016, background): “We refer the reader to [39] for a comparison of specific”
Metrics Reloaded: Recommendations for image analysis validation (2022, unknown): “Despite their importance, an increasing body of work shows that the metrics used in common practice often do not adequately reflect the underlying biomedical problems, diminishing the validity of the investigated algorithms [31, 36, 41, 52, 54, 58, 62, 88].”
Structure-Measure: A New Way to Evaluate Foreground Maps (2017, methodology): “ Recall curve by plotting Precision p(r) as a function of Recall r. AP measure [17] is the average value of p(r) over the evenly spaced x axis points from r= 0 to r= 1. Recently, a measure called Fbw [36] has offered an intuitive generalization of the F -measure. It is deﬁned as: F! = (1 + 2)Precisi”
PiCANet: Learning Pixel-Wise Contextual Attention for Saliency Detection (2017, methodology): “We also follow [5, 34, 10] to adopt it as one of our metrics with the default settings in [27].”
Camouflaged Object Detection (2020, background): “Recent studies [12, 13] have suggested that the weighted F-measure ( F wβ ) [43] can provide more reliable evaluation results than the traditional F β ; thus, we also consider this metric in the COD ﬁeld.”
Stacked Cross Refinement Network for Edge-Aware Salient Object Detection (2019, background): “Then weighted F-measure ( F ωβ ) [25] and structural similarity measure ( S α , α = 0 .”
What Do Different Evaluation Metrics Tell Us About Saliency Models? (2016, background): “ntrol model, which has more zero values, fares worse than the chance model. One possible meta-measure for selecting metrics for evaluation is how much better one baseline is over another (e.g., [23], [56], [65]). However, the optimal ranking of baselines is likely to be different across applications”
AffordanceNet: An End-to-End Deep Learning Approach for Object Affordance Detection (2017, methodology): “Baseline As the standard practice, we use the F β metric [26] to evaluate the affordance detection results.”
Multi-Scale Interactive Network for Salient Object Detection (2020, background): “…pixel values with the image-level mean value to jointly evaluate the similarity between the prediction and the ground truth. weighted F-measure [25], F ωβ deﬁnes a weighted precision, which is a measure of exactness, and a weighted recall, which is a measure of completeness is proposed to…”
Automatic Polyp Segmentation via Multi-scale Subtraction Network (2021, methodology): “We adopt several widely used metrics for quantitative evaluation: mean Dice, mean IoU, the weighted F-measure (F β ) [18], mean absolute error (MAE), the recently released S-measure (Sα) [5] and E-measure (E max φ ) [6] scores.”
Salient Object Detection via Structured Matrix Decomposition (2017, methodology): “, precision-recall curve and mean absolute error, and the recently proposed weighted F measure [32].”
Camouflaged Object Segmentation with Distraction Mining (2021, methodology): “We use four widely used and standard metrics to evaluate our method: structure-measure (Sα) [10], adaptive E-measure (E φ ) [11], weighted F-measure (F β ) [35], and mean absolute error (M ).”
Camouflaged Object Detection with Feature Decomposition and Edge Reconstruction (2023, methodology): “Four commonly-used metrics are employed for COD task, including mean absolute error M , adaptive F-measure Fβ [27], mean E-measure Eφ [5], and structure measure Sα [3].”
Salient Object Detection in the Deep Learning Era: An In-Depth Survey (2019, methodology): “…used to remedy this problem by measuring the average pixel-wise absolute error between normalized map S ∈ [0 , 1] W × H and ground-truth mask G ∈{ 0 , 1 } W × H : (3) • Weighted F β measure (Fbw) [127] intuitively generalizes F-measure by alternating the way to calculate the Precision and Recall.”
U2-Net: Going Deeper with Nested U-Structure for Salient Object Detection (2020, methodology): “To comprehensively evaluate the quality of those probability maps against the ground truth, six measures including (1) Precision-Recall (PR) curves , (2) maximal Fmeasure (maxFβ) [1] , (3) Mean Absolute Error (MAE) [23, 33, 22], (4) weighted F-measure (F β ) [29] , (5) structure measure (Sm) [8] and”
Polyp-PVT: Polyp Segmentation with Pyramid Vision Transformers (2021, methodology): “We employ six widely-used evaluation metrics, including Dice [85], IoU, mean absolute error (MAE), weighted Fmeasure (F β ) [86], S-measure (Sα) [87], and E-measure (Eξ) [88], [89] to evaluate the model performances.”