This dataset contains leaf venation architecture traits for 27 angiosperm tree species from a tropical forest-savanna rainfall gradient in Ghana, West Africa. Collection sites comprised evergreen forest (Ankasa; 5.26°N, 2.69°W; mean annual precipitation [MAP] 2000 mm), semi-deciduous forest (Bobiri; 6.69°N, 1.32°W; MAP 1500 mm), and dry forest (Kogyae; 7.26°N, 1.15°W; MAP 1200 mm). 1-2 leaves per species were chemically cleared, imaged with a high-resolution flatbed scanner, and processed using LeafVeinCNN software in Matlab. This dataset comprises the cleared leaf images, venation segmentations, extracted networks (spatial graphs), and venation architectural traits generated from LeafVeinCNN. This dataset is useful for leaf venation architectural comparison between plant species and across rainfall gradients.