Page 60 - 2025S

UEC Int’l Mini-Conference No.54 53

Figure 3: Evaluate the efficacy of the trained DenseNet201 model for automated medicinal plant
identification based on leaf

lection methods such as EFS were also employed 3 Methodology
in recent models [4], which gained popularity
due to the depth and complexity of modern ar- 3.1 Data acquisition
chitectures. Table 1 presents a comparison and This study utilized leaf images from six
gap analysis of existing works. In conclusion, Bangladeshi medicinal plant species, collected
based on an exhaustive literature review, we
from local fields in Ashulia, Savar, Dhaka. Tra-
found that no automated application currently ditional herbal plants are utilized for enhancing
exists in Bangladesh to assist the general public immunity [30]. Fig.1 illustrates the medicinal
in accurately identifying medicinal plants. To plant identification workflow.
address this gap, this work focuses on develop-
ing an automated application capable of detect-
ing medicinal plants without relying on human 3.2 Data Pre-Processing
visual identification.
This study employs comprehensive pre-
processing techniques, including file format op-
timization, histogram equalization, background
removal, data cleaning, data augmentation, and
gamma correction.