Unsupervised anomaly detection in time-series has been extensively investigated in the literature. Notwithstanding the relevance of this topic in numerous application fields, a complete and extensive evaluation of recent state-of-the-art techniques is still missing. Few efforts have been made to compare existing unsupervised time-series anomaly detection methods rigorously. However, only standard performance metrics, namely precision, recall, and F1-score are usually considered. Essential aspects for assessing their practical relevance are therefore neglected. This paper proposes an original and in-depth evaluation study of recent unsupervised anomaly detection techniques in time-series. Instead of relying solely on standard performance metrics, additional yet informative metrics and protocols are taken into account. In particular, (1) more elaborate performance metrics specifically tailored for time-series are used; (2) the model size and the model stability are studied; (3) an analysis of the tested approaches with respect to the anomaly type is provided; and (4) a clear and unique protocol is followed for all experiments. Overall, this extensive analysis aims to assess the maturity of state-of-the-art time-series anomaly detection, give insights regarding their applicability under real-world setups and provide to the community a more complete evaluation protocol.