Semantic segmentation, or image segmentation, is the task of clustering parts of an image together which belong to the same object class. It is a form of pixel-level prediction because each pixel in an image is classified according to a category (Thoma, 2016).
The PASCAL-Context dataset is a challenging scene parsing dataset that contains 59 semantic classes and 1 background class (Mottaghi et al., 2014). The training set and test set consist of 4, 998 and 5,105 images respectively.
An excellent reference for tracking state-of-the-art models is PapersWithCode, which tracks performance data of ML models.
What will the state-of-the-art performance on semantic segmentation of PASCAL-Context be on 2023-02-14 in mean IoU in percent (MIoU%), amongst models not trained on extra data?
This question resolves as the highest level of performance (in MIoU%) achieved on the PASCAL-Context (2014) dataset up until 2023-02-14, 11:59PM GMT amongst models trained on only the PASCAL-Context training set—no extra training data may be used. The model's level of performance is to be evaluated on the PASCAL-Context test set.
Performance figures may be taken from e-prints, conference papers, peer-reviewed articles, and blog articles by reputable AI labs (including the associated code repositories). Published performance figures must be available before 2023-02-14, 11:59PM GMT to qualify.