This dataset supports a study in which a ML image classification model was developed to automate the scoring of the Test of Creative Thinking - Drawing Production (TCT-DP). The TCT-DP is a figural test of creativity in which individuals draw on a pre-defined test form, not to be confused with the Torrance Tests of Creativity (TTCT). Note: the TCT-DP is frequently confused with the TTCT! Three sets of scored TCT-DP images were used to train the ML model. Dataset 1 comprises 5 subsets (K1, K2, K4, K5, K6) and consists of a total of approx. 1,800 original (unrotated) images. Dataset 2 consists of approximately 350 original (unrotated) images, and Dataset 3 consists of approximately 240 original (unrotated) images. In each dataset, the images are grouped in folders according to their broad classification, using the scoring bands for 11-16 year olds, as defined in the TCT-DP test manual. In other words: Far Below Average (=score <16) Below Average (=score 16-20) Average (=score 21-33) Above Average (=score 34-40) Far Above Average (=score 41+) In addition, the images are stored both in their "original" (unrotated) form - i.e., the test form image is saved in the normal, vertical orientation in which it is presented to test-taker - and also with three rotations (90, 180 and 270 degrees). These rotated augmentations can be used supplement the original, unrotated images, as appropriate. The dataset also includes an "external validation" set of images. This is an additional set of 79 TCT-DP images that were NOT used in the creation of the ML model (for any part of training, validation or testing). This allows for additional testing of the model on top of any internal validation and testing, and guarantees that there is no possibility of data leakage. A breakdown of the numbers of images and different classes is included in the attached Excel spreadsheet. The current folder structure of the data reflects GCP Model 2.