Poisonous Plant Prediction Using Explainable Deep Inherent Learning Model

Area of Science:

• Computer Vision
• Artificial Intelligence
• Botany

Background:

• Global plant species discovery presents challenges in distinguishing beneficial from poisonous varieties.
• Existing computer vision methods for plant classification and toxicity prediction are limited by data availability.
• Lack of comprehensive datasets hinders accurate identification and prediction of poisonous plants.

Purpose of the Study:

• To propose an Explainable Deep Inherent Learning approach for plant species classification and poisonous status prediction.
• To enhance the accuracy and interpretability of AI models in identifying poisonous plants.
• To build trust in AI-driven systems for plant identification and safety.

Main Methods:

• Leveraged advanced computer vision techniques within a Deep Inherent Learning framework.
• Employed Explainable AI (XAI) for local and global decision-making process clarification.
• Validated the approach using a dataset of 2500 images from 50 Arabian Peninsula plant species with metadata.

Main Results:

• The XAI model achieved high performance metrics: 0.94 accuracy, 0.96 Precision, 0.96 Recall, and 0.97 F1-Score.
• Demonstrated effective plant species classification and poisonous status prediction.
• Provided visual information to enhance user trust in the AI model's predictions.

Conclusions:

• The proposed Explainable Deep Inherent Learning approach is effective for plant classification and toxicity prediction.
• Enhanced interpretability through XAI fosters greater trust in AI for identifying harmful plants.
• This research contributes to reducing poisoning incidents from harmful plants, benefiting public health.