Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

GRITv2: Efficient and Light-weight Social Relation Recognition

N K Sagar Reddy, Neeraj Kasera, Avinash Thakur

TL;DR

The proposed GRITv2-L surpasses existing methods on relation recognition and the GRITv2-S is within 2% performance gap of GRITv2-L, which has only 0.0625x the model size and parameters of GRITv2-L.

Abstract

Our research focuses on the analysis and improvement of the Graph-based Relation Inference Transformer (GRIT), which serves as an important benchmark in the field. We conduct a comprehensive ablation study using the PISC-fine dataset, to find and explore improvement in efficiency and performance of GRITv2. Our research has provided a new state-of-the-art relation recognition model on the PISC relation dataset. We introduce several features in the GRIT model and analyse our new benchmarks in two versions: GRITv2-L (large) and GRITv2-S (small). Our proposed GRITv2-L surpasses existing methods on relation recognition and the GRITv2-S is within 2% performance gap of GRITv2-L, which has only 0.0625x the model size and parameters of GRITv2-L. Furthermore, we also address the need for model compression, an area crucial for deploying efficient models on resource-constrained platforms. By applying quantization techniques, we efficiently reduced the GRITv2-S size to 22MB and deployed it on the flagship OnePlus 12 mobile which still surpasses the PISC-fine benchmarks in performance, highlighting the practical viability and improved efficiency of our model on mobile devices.

GRITv2: Efficient and Light-weight Social Relation Recognition

TL;DR

The proposed GRITv2-L surpasses existing methods on relation recognition and the GRITv2-S is within 2% performance gap of GRITv2-L, which has only 0.0625x the model size and parameters of GRITv2-L.

Abstract

Our research focuses on the analysis and improvement of the Graph-based Relation Inference Transformer (GRIT), which serves as an important benchmark in the field. We conduct a comprehensive ablation study using the PISC-fine dataset, to find and explore improvement in efficiency and performance of GRITv2. Our research has provided a new state-of-the-art relation recognition model on the PISC relation dataset. We introduce several features in the GRIT model and analyse our new benchmarks in two versions: GRITv2-L (large) and GRITv2-S (small). Our proposed GRITv2-L surpasses existing methods on relation recognition and the GRITv2-S is within 2% performance gap of GRITv2-L, which has only 0.0625x the model size and parameters of GRITv2-L. Furthermore, we also address the need for model compression, an area crucial for deploying efficient models on resource-constrained platforms. By applying quantization techniques, we efficiently reduced the GRITv2-S size to 22MB and deployed it on the flagship OnePlus 12 mobile which still surpasses the PISC-fine benchmarks in performance, highlighting the practical viability and improved efficiency of our model on mobile devices.
Paper Structure (25 sections, 10 equations, 2 figures, 8 tables)

This paper contains 25 sections, 10 equations, 2 figures, 8 tables.

Table of Contents

  1. Introduction
  2. Related Works
  3. Social Relation Recognition
  4. Graph Neural Networks
  5. GRIT
  6. Feature Extraction Module
  7. Graph-based Query Module
  8. Transformer Reasoning Module
  9. GRITv2
  10. Weighted Binary Cross Entropy
  11. Bilateral Masking
  12. Logit Transformation
  13. GQM Update
  14. Squeeze and Excitation Block
  15. Comparision with SOTA
  16. ...and 10 more sections

Figures (2)

  • Figure 1: Architecture of GRITgrit, consisting of 3 modules: FEM, GQM and TRM.
  • Figure 2: PISC-CPISC_dualglance and PISC-FPISC_dualglance train split analysis. Int: Intimate, Non: Non-Intimate, NoR: No Relation, Fri: Friend, Fam: Family, Cou: Couple, Pro: Professional, Com: Commercial.