Visualizing the Effect of Semantic Categories in the Attribution of Scene Recognition Models

Alejandro López-Cifuentes, Marcos Escudero-Viñolo, Andrija Gajic and Jesús Bescós

Video Processing and Understanding Lab (VPU Lab), Universidad Autónoma de Madrid

Official project site of Visualizing the Effect of Semantic Classes in the Attribution of Scene Recognition Models (ICPR 2020 EDL-AI Workshop).

Abstract

The performance of Convolutional Neural Networks for image classification has vastly and steadily increased during the last years. This success goes hand in hand with the need to explain and understand their decisions: opening the black box.

The problem of attribution specifically deals with the characterization of the response of Convolutional Neural Networks by identifying the input features responsible for the model's decision. Among all attribution methods, perturbation-based methods are an important family based on measuring the effect of perturbations applied to the input image in the model's output. In this paper, we discuss the limitations of existing approaches and propose a novel perturbation-based attribution method guided by semantic segmentation.

Our method inhibits specific image areas according to their assigned semantic label. Hereby, perturbations are link up with a semantic meaning and a complete attribution map is obtained for all image pixels. In addition, we propose a particularization of the proposed method to the scene recognition task which, differently than image classification, requires multi-focus attribution models. The proposed semantic-guided attribution method enables us to delve deeper into scene recognition interpretability by obtaining for each scene class the sets of relevant, irrelevant and distracting semantic labels.

Experimental results suggest that the method can boost research by increasing the understanding of Convolutional Neural Networks while uncovering datasets biases which may have been inadvertently included during the harvest and annotation processes.

Site Contents

Code

Github link for the available code to reproduce the results of the paper.

Supplementary Material

Contains complete semantic statistics for the 365 classes from Places-365 dataset. Each scene class contains:

  • Relevant Semantics
  • Irrelevant Semantics
  • Distracting Semantics
  • Semantic Distributions

Dataset

Contains the used semantic segmentation images for the validation set of Places365. Each semantic image contains 3 labels per pixel (Top@3) using the 3 image channels. Paper results used Top@1 prediction which is encoded in the third image channel.

Citation

If you find this work useful, please consider citing:

López-Cifuentes, A., Escudero-Viñolo, M., Gajic A., & Bescós, J. (2021). Visualizing the Effect of Semantic Classes in the Attribution of Scene Recognition Models. In Proceedings of Pattern Recognition. ICPR International Workshops and Challenges. (2021)

 
  @InProceedings{Lopez2020Visualizing,
    author="L{\'o}pez-Cifuentes, Alejandro and Escudero-Vi{\~{n}}olo, Marcos and Gaji{\'{c}}, Andrija and Besc{\'o}s, Jes{\'u}s",
    title="Visualizing the Effect of Semantic Classes in the Attribution of Scene Recognition Models",
    booktitle="Pattern Recognition. ICPR International Workshops and Challenges",
    year="2021",
    pages="115--129",
  }
   				

Acknowledgement: This study has been partially supported by the Spanish Government through its TEC2017-88169-R MobiNetVideo project.