Flower image recognition has gained attention due to its relevance in botany, agriculture, and biodiversity studies. This project employs the Inception-v4 model, a deep convolutional neural network architecture, for classifying flower species using the Oxford 17 Category Flower Dataset. Transfer learning was used to enhance model generalization, while data augmentation was applied to address overfitting. The model achieved a test accuracy of 83.33%, demonstrating robust classification performance. Additionally, a user-friendly GUI was developed using PyQt5, allowing users to upload flower images for real-time classification.
The integration of Inception-v4 and PyQt5 in this project highlights the potential of deep learning in real-world flower recognition systems. Compared to traditional classification approaches, the model significantly improves accuracy and usability, offering a practical solution for plant identification in research and field settings.
The model is limited by the dataset size and class imbalance. Generalization to unseen flower species may be reduced without further training. The GUI, while functional, lacks mobile compatibility and multilingual support.
Future improvements include training on larger, multi-institutional flower datasets, integrating multilingual support, and deploying the system as a web-based app or mobile application using TensorFlow Lite or ONNX.