One of the critical issues faced by a Koi's breeder is sexing at the early age, since it will result in financial benefit for the breeder. In this paper we describe the results of early sexing technique that is developed for the Koi using combination of descriptor method and application of artificial neural network (ANN), where we used back-propagation algorithm in the computation. Input data for ANN are obtained from both morphometric and energetic descriptors of Koi's images. The morphometric descriptor shows that early sexing is possible and can be accomplished within 2 (two) months, which is much faster than conventional technique by means of histology. The accuracy of ANN computation for sexing using both descriptors as input is about 70%.