Improving The Germination Quality Assessment of Rice Seeds Using Image Annotation and Deep Learning Models

Rice is a staple food crop that sustains more than half of the global population, and the quality of rice seeds plays a decisive role in determining-‎ing germination success, crop establishment, and eventual yield. Accurate assessment of germination quality is therefore essential for ensuring seed performance in both breeding and commercial farming systems. Conventional germination assessment methods, such as germination percentage tests and vigor evaluations, are labor-intensive, time-consuming, and often destructive, making them unsuitable for large-scale or real-time monitoring. To overcome these limitations, this study proposes a deep learning-based framework for non-destructive ger-‎germination quality assessment using annotated rice seed imagery. The Germination Seed Quality dataset from Roboflow, comprising 2,069 ‎labeled images of rice seeds, was employed for model development. After preprocessing and data augmentation, a convolutional neural ‎network was trained to classify seeds based on germination quality. Experimental results demonstrate that the proposed model achieved an ‎overall classification accuracy of 94.7%, with a precision of 94.1%, recall of 93.8%, and an F1-score of 0.93, outperforming baseline ma-‎machine learning approaches. Visualization of activation maps further revealed the morphological features most critical for model predictions, ‎providing interpretability of the decision process. These findings highlight the potential of deep learning and annotated imagery as a scalable, ‎efficient, and non-destructive solution for rice seed germination quality monitoring. The proposed approach holds promise for integration ‎into precision agriculture pipelines, enabling seed producers, breeders, and farmers to make informed decisions that improve productivity ‎and sustainability‎.