Jasminum Grandiflorum flower images classification: deep learning and transfer learning models with the influence of preprocessing via contours and convex hull in Agritech 4.0

This study specifically centres on classifying Jasminum Grandiflorum flowers through the utilisation of deep learning and transfer learning techniques. To achieve this, the research leverages advanced deep learning models such as CNNs, along with transfer learning using pre-trained architectures like VGG16, VGG19, ResNet18, and Vision Transformer. CNN stood out, excelling after extensive iterations. VGG 16 and 19 showed solid performance with fewer iterations, indicating competence in shorter training times. ResNet18 achieved the highest accuracy with fewer iterations but took longer (about 8 minutes per epoch), balancing efficiency and accuracy. ViT impressed with high accuracy despite needing more iterations, showcasing prowess in intricate learning and pattern recognition in the Jasminum Grandiflorum flower image dataset. The intended outcome of this research is to contribute significantly to the advancement of Agritech 4.0 by establishing a robust methodology for accurate Jasminum Grandiflorum flower classification without human participation.