Article Per Year (5 Year)
p-Index From 2021 - 2026
0.23
P-Index
This Author published in this journals
All Journal Proceeding of International Conference Health, Science And Technology (ICOHETECH)
Kevien Ollyvie Jolanda
Universitas Duta Bangsa Surakarta
Computer Science & IT Decision Sciences, Operations Research & Management Health Professions Industrial & Manufacturing Engineering Medicine & Pharmacology

Published : 1 Documents Claim Missing Document
Claim Missing Document
Check
Articles

Found 1 Documents
Search

 1 
Diagnosis and Mitigation of INT8 Quantization Failure in MobileNetV3 for Plant Leaf Disease Classification Sopingi; Afu Ichsan Pradana; Kevien Ollyvie Jolanda; Nauval Nabil Iman
Proceeding of the International Conference Health, Science And Technology (ICOHETECH) 2026: Proceeding of the 7th International Conference Health, Science And Technology (ICOHETECH)
Publisher : LPPM Universitas Duta Bangsa Surakarta

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.47701/qs6esj44

Abstract

Edge deployment of deep learning models for plant disease classification requires model compression to fit resource-constrained hardware, yet post-training quantization (PTQ) may introduce accuracy loss that is not always well characterized for compact architectures. This progress report presents interim results from a wider study optimizing a fine-tuned MobileNetV3-Small model for tomato leaf disease classification. After training reached 99.71% validation accuracy on a 10-class PlantVillage subset, standard INT8 static PTQ using ONNX Runtime and TensorFlow Lite produced a severe accuracy collapse (10.4-27.4%) despite systematically testing eleven quantization configurations, including per-channel/per-tensor granularity, calibration methods, operator exclusion, and export settings. Diagnostic analysis revealed a persistent class-wise failure pattern rather than random noise, and an explicit shape-mismatch error confirmed a quantizer defect in per-channel bias requantization for grouped convolutions. As a mitigation, dynamic weight-only and dynamic-range quantization schemes were adopted, restoring laboratory accuracy to 99.4% with 29.3-72.0% size reduction, although field-condition accuracy on secondary open field data dropped to 43.1-43.9%, indicating a substantial domain generalization gap that motivates the next research phase.
Co-Authors Afu Ichsan Pradana Nauval Nabil Iman Sopingi