Body condition score (BCS) is a proxy for evaluating body fat reserves. However, monitoring BCS is a time-consuming and subjective task. Thus, we aimed to develop a method for estimating the BCS of beef cows using three-dimensional (3D) body features of cows’ rump area derived from 3D camera data. Three-dimensional surface data of the rump area from 39 multiparous cows were obtained using a 3D camera, and four 3D body features were extracted. The BCSs of the cows were scored by experts, and models for predicting BCS by 3D features were developed using machine learning algorithms. The derived model yielded an overall accuracy, precision, sensitivity, and F-measure of 90%, 88%, 90%, and 88%, respectively. Additionally, we evaluated a simple practical method to estimate BCS using the difference between heart girth (HG) and tightened heart girth (THG) for 118 multiparous cows. A cumulative logistic regression model for estimating BCS by the difference was developed, and the derived generalized coefficient of determination was 0.81. These results suggest that 3D images are useful for estimating the BCS of beef cows and that the difference between HG and THG can be used to estimate BCS as a simple practical method.