Online monitoring of cell density and cell viability is a challenging but essential task to control and optimize biotechnical processes and is of particular interest for the growing field of animal cell cultures. For this purpose, we introduce an optical approach for automated cell detection and viability classification of suspended mammalian cells. Our proposed system CellViCAM is capable of evaluating dark field micrographs by means of several image processing and supervised machine learning techniques without the use of any dyes or fluorescent labeling. Using a human cell line as the reference culture, an efficient cell detection procedure has been established also enabling a cell density estimation. Furthermore, a comprehensive but reagent-free viability analysis, based on a semi-automatic training data generation, has been developed. By means of an extensive validation dataset we can show that the CellViCAM approach can be considered as an equivalent to staining-based methods and moreover, how it provides a technical platform for a more differentiated cell state classification into living, necrotic, early and late apoptosis.