Submodular Field Grammars: Representation, Inference, and Application to Image Parsing

Authors: Abram L. Friesen, Pedro M. Domingos

NeurIPS 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Empirically, we show promising improvements in accuracy when using SFGs for scene understanding, and demonstrate exponential improvements in inference time compared to traditional methods, while returning comparable minima.
Researcher Affiliation Academia Abram L. Friesen and Pedro Domingos Paul G. Allen School of Computer Science and Engineering University of Washington Seattle, WA 98195 {afriesen,pedrod}@cs.washington.edu
Pseudocode Yes Algorithm 1 Compute the (approximate) MAP parse of an image with respect to an SFG.
Open Source Code No The paper mentions supplementary material but does not provide a direct link to source code or explicitly state that source code for the methodology is released.
Open Datasets Yes The Deep Lab features are trained on the Stanford Background Dataset (SBD) [5] training set.
Dataset Splits No The paper mentions a 'training set' and 'test set' but does not explicitly describe a validation set or its specific split for reproduction.
Hardware Specification Yes The DGX-1 used for this research was donated by NVIDIA.
Software Dependencies No The paper mentions the use of 'Deep Lab features' and algorithms like 'belief propagation' and 'α-expansion' but does not provide specific version numbers for software dependencies or libraries.
Experiment Setup Yes In all MRFs (planar and in each SFG), the pairwise terms are standard contrast-dependent boundary terms [25] multiplied by a single weight, w BF.