Ilex lanceifolia K.W.Xu & Lei Jiang, a new species from the western Pearl River Delta of Guangdong, China, is described based on morphological and molecular evidence. To test whether this newly discovered population represents a distinct lineage and to assess the congruence between leaf morphology and phylogeny, we integrated multivariate morphometrics, scanning electron microscopy, and phylogenetic analyses of nuclear ITS, ETS, and nepGS sequences. The new species resembles I. xiaojinensis and I. peiradena in shrubby habit and lanceolate leaves but differs by its prominently raised abaxial leaf veins forming distinct reticulate areoles, and green to purplish-black petioles and young branchlets. However, phylogenetic analyses unexpectedly place it within Ilex sect. Ilex forming a clade with I. graciliflora and six other species, rather than with its morphological look-alikes. This discordance strongly suggests that the lanceolate leaf shape has evolved convergently in multiple lineages of Ilex, likely as an adaptive strategy to the high-humidity, low-light understory conditions of subtropical lowland forests. The new species is currently known only from a single population in Jiangmen City, with several thousand individuals but an extremely restricted range (<20 km²), warranting conservation attention. This discovery highlights the underestimated biodiversity of lowland forests in the Pearl River Delta and underscores the need to prioritize remnant habitat fragments in rapidly urbanizing regions.

Holly (Ilex L.) is a genus of woody dioecious plants with substantial ecological and economic value. However, its high species diversity and morphological similarity make accurate identification challenging. To address this, we constructed a multi-taxon Ilex leaf image dataset. We then trained six deep learning models—GoogLeNet, ResNet50, ResNet101, DenseNet121, DenseNet169, and EfficientNet-B3—using a unified PyTorch framework on cloud computing resources. Leaf images were preprocessed by background removal, resizing, cropping, and normalization. Model performance was evaluated using accuracy, F1-score, and Grad-CAM visualizations. Under an image-level data split that may overestimate generalization, all six models achieved over 99% classification accuracy on preprocessed leaf images under controlled laboratory conditions. DenseNet121 and DenseNet169 performed best, reaching 99.65% accuracy. Because images of the same leaf or same plant could appear in both training and test sets under this split, plant-level cross-validation is required to assess real-world generalizability. The reported accuracies represent an upper-bound estimate under image-level splitting. The framework offers a rapid and accurate tool for preliminary screening under controlled conditions, but its performance on raw field photographs and across different collection sites remains to be validated.