Dipashree Sadhukhan

Dipashree Sadhukhan

Auto-generated profile, based on publicly available information

Dipashree Sadhukhan

Dept. of Plant Pathology, Bidhan Chandra Krishi Viswavidyalaya

Research Focus & Metrics

Academic Career Length

2Years (2025-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
24 July 2025 in Archives of Phytopathology and Plant Protection
Verified
Journal Article
Closed access
ANAnanda Nath
Ananda Nath
SRSoumyayan Roy
Soumyayan Roy
DSDipashree Sadhukhan
Dipashree Sadhukhan
SBSudip Bhattacharya
Sudip Bhattacharya
SMSabyasachi Mukhopadhyay
Sabyasachi Mukhopadhyay
SPSushil Pandey
Sushil Pandey

3 October 2026 in Plant Science Today
Verified
Journal Article
Open access
PDF file
PSP Soumen
P Soumen
MSM Sabyasachi
M Sabyasachi
SAS Ayan
S Ayan
SDS Dipasree
S Dipasree
APA Payel
A Payel
SDSubrata Dutta
Subrata Dutta
Rice (Oryza sativa L.) constitutes a primary food source for a significant proportion of the global population and its grain quality directly influences market value and consumer acceptance. Grain discolouration (GD), largely attributable to seed-borne pathogens and occasionally to insect infestation, poses a critical challenge by compromising visual quality and reducing commercial grade. This study evaluated the performance of advanced deep learning models for automated detection and classification of discoloured rice grains, incorporating criteria related to predictive accuracy, computational requirements and environmental sustainability. Among the assessed architectures, DenseNet121 exhibited the highest accuracy (98 %), followed by MobileNet (96 %) and MobileNetV2 (92 %), each demonstrating strong precision, recall and F1 scores. In contrast, VGG19 recorded comparatively lower accuracy (84 %). Computational and energy analysis revealed that DenseNet121, MobileNet and Xception required shorter training durations and lower computational resources, whereas MobileNetV2 and InceptionV3 were comparatively resource-intensive. Notably, MobileNet demonstrated superior energy efficiency, consuming 19.03 watts and generating 0.00952 kg carbon dioxide (CO₂), whereas InceptionV3 consumed 377.27 watts and emitted 0.18864 kg CO₂. To facilitate comprehensive model prioritisation, the technique for order of preference by similarity to ideal solution (TOPSIS) was employed, identifying MobileNet as the optimal model when considering performance, efficiency and environmental impact. The findings highlight the potential applicability of lightweight deep learning architectures, particularly MobileNet and DenseNet121, in real-time rice grading and sorting systems, supporting sustainable post-harvest management, improving grain quality assurance and mitigating seed-borne pathogen dissemination.