Abstract:
A custom dataset of Neem leaves and trees was collected and categorized into multiple
age and health classes, forming the basis for developing an automated deep learningbased classification framework. Since Neem is a tree of high medicinal, ecological and
cultural importance, accurate identification of its leaf and tree age can contribute to
better utilization of its properties and support ecological conservation. A series of
preprocessing steps were performed to ensure that the dataset was suitable for deep
learning applications . These included preprocessing properly the images to a
consistent resolution , normalizing pixel values to bring them within a standard range
and applying extensive data augmentation techniques such as rotations , flips , scaling
and color modifications . For the image classification task , transfer learning was used
on several pretrained convolutional neural network (CNN) architectures , including
DenseNet121 , ResNet50 , MobileNet , NASNet and EfficientNet . Transfer learning
allowed the models to leverage previously learned feature representations from largescale image datasets like ImageNet while being fine-tuned to the specific
characteristics of Neem leaf and tree images . By giving an automated and reliable
method of Neem leaf and tree age classification , the framework reduces the
dependency on manual observation , which is often subjective and error-prone , and
instead provides a proper solution that can be supportive for medicinal research ,
ecological monitoring and resource optimization , ultimately contributing to both
scientific advancement and environmental sustainability .
