With post-mortem iris recognition getting increasing attention throughout the biometric and forensic communities, no specific, cadaver-aware recognition methodologies have been proposed to date. This paper makes the first step in assessing the discriminatory capabilities of post-mortem iris images collected in multiple time points after a person's demise, by proposing a deep convolutional neural network (DCNN) classifier fine-tuned with cadaver iris images. The proposed method is able to learn these features and provide classification of post-mortem irises in a closed-set scenario, proving that even with post-mortem biological processes' onset after a person's death, features in their irises remain, and can be utilized as a biometric trait. This is also the first work (known to us) to analyze the class-activation maps produced by the DCNN-based iris classifier, and to compare them with attention maps acquired by a gaze-tracking device observing human subjects performing post-mortem iris recognition task. We show how humans perceive post-mortem irises when challenged with the task of classification, and hypothesize that the proposed DCNN-based method can offer human-intelligible decisions backed by visual explanations which may be valuable for iris examiners in a forensic/courthouse scenario.