Artificial Intelligence-based data analytics is widely adopted to automate variety of classification tasks. Recently, the deep learning (DL)-based image classification is commonly adopted in various domains to achieve better detection accuracy. This work considered the rice variety detection as the task and to achieve the accuracy, this work proposed a DL-scheme. This tool consist the following phases; (i) image collection and resizing, (ii) feature extraction using a chosen DL-model, (iii) feature reduction and serial features fusion, and (iv) classification with threefold cross-validation. In this work, the identification of the rice variety using its digital image is considered as the task and the proposed experiment is executed using the Python software. This work employed the ResNet-variants to detect the appropriate rice variety using the SoftMax classifier. The experimental outcome of this study confirms that the implemented approach provides a detection accuracy of > 91% with the conventional ResNet-feature and > 97% with the fused ResNet-features. The result confirms that the proposed scheme provides a significant result for the chosen task.