Real Time YOLOv8 Training Monitor: A Lightweight Diagnostic Tool with Overfitting Detection and Remote Alerting

View at publisher Summarise with VIRAT
Authors: M. Indirapriyadharshini, Muhammed Roshan S, Kiruthiga A, Ragul S J
Cited by: 0
References: 6

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

YOLOv8 is a prominent object detection framework recognized for combining real-time processing speed with strong detection accuracy. However, when trained on limited or imbalanced datasets, it may still encounter overfitting, leading to weak generalization and inefficient use of computational resources. To address this, we introduce a lightweight real-time monitoring tool that dynamically visualizes both training and validation loss metrics. The system detects overfitting patterns, such as a consistent increase in validation loss over multiple epochs, and can issue automatic alerts via the Telegram Bot API. When overfitting persists for five or more epochs, it may halt training to prevent further resource waste. This approach reduces unnecessary GPU consumption, improves model reliability, and integrates seamlessly into existing YOLOv8 workflows with minimal setup cost.

Fields of Study & Subjects

Advanced Neural Network Applications Machine Learning and Data Classification Smart Agriculture and AI

Publication details

Journal / source
International Conference on Intelligent Systems for Communication, IoT and Security (ICISCoIS)
Pages
1–5
Publication date
8 January 2026
Language
EN