# A Low-Shot Object Counting Network With Iterative Prototype Adaptation

**Authors:** Nikola Djukic, Alan Lukezic, Vitjan Zavrtanik, Matej Kristan  
**arXiv:** [2211.08217](https://arxiv.org/abs/2211.08217)  
**Date:** 2022-11-15  
**Summary page:** https://www.summarizepaper.com/en/paper/2211.08217v2/

## Key points

- Introduction of a novel framework called LOCA for object counting in computer vision
- Incorporation of shape information through an unspecified method to address limitations of existing few-shot pipelines
- Outperforms state-of-the-art techniques in terms of localization and count estimation accuracy
- Suitable for a wide range of scenarios including zero-shot scenarios
- Potential future research directions include introducing additional supervision levels and narrowing the gap between low-shot counters and object detectors
- LOCA offers a promising solution for accurate and efficient object counting in computer vision

## Summary

, , , , 

The paper introduces a novel framework, called LOCA, for . It addresses the limitations of existing few-shot pipelines by incorporating shape information through an . The proposed method outperforms state-of-the-art techniques in terms of both localization and count estimation accuracy, making it suitable for a wide range of scenarios including <kd> zero-shot scenarios</ kd>. The authors also discuss potential future research directions such as introducing additional supervision levels and narrowing the gap between low-shot counters and object detectors. Overall, LOCA offers a promising solution for accurate and efficient object counting in computer vision.