Evaluation Metrics for CNNs Compression

CoRR(2023)

引用 0|浏览8
暂无评分
摘要
There is a lot of research effort devoted by researcher into developing different techniques for neural networks compression, yet the community seems to lack standardised ways of evaluating and comparing between different compression techniques, which is key to identifying the most suitable compression technique for different applications. In this paper we contribute towards standardisation of neural network compression by providing a review of evaluation metrics. These metrics have been implemented into NetZIP, a standardised neural network compression bench. We showcase some of the metrics reviewed using three case studies focusing on object classification, object detection, and edge devices.
更多
查看译文
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络