Cascade detectors have been widely used in several applications such as face recognition and pedestrian detection because they are fast and provide good results. Soft cascades are a variant of the classic cascade detectors. This new type of cascades is implemented in OpenCV 3 in the softcascade module. Soft cascade is trained with AdaBoost but the resulting detector is composed of only one stage. This stage has several weak classifiers that are evaluated in sequence. After evaluating each weak classifier, the result is compared with the corresponding threshold. Similar to the evaluation process carried out in multistage cascades, negative non-object instances are discarded as soon as possible.