Imaging-based spatial genomics can provide unprecedented detail on the cellular architecture of complex tissues. Recent advances enable highly-multiplexed gene expression measurements in situ and promise a 3D molecular map of human tissues at the single cell resolution. However, mapping entire human tissues at single-cell quantitative resolution presents significant challenges in imaging, data analysis and dissemination. I will talk about our previous work on developing a large-scale automated spatial transcriptomics pipeline that combines single molecule fluorescent in situ hybridization (smFISH) and fast high-resolution confocal microscopy. I will present the application of this pipeline to mapping the cellular diversity in the mouse cerebral cortex. Last, I will present our current work on the Human Cell Atlas (HCA) that aims to combine highly-multiplexed smFISH with high-throughput imaging and deliver a spatially resolved molecular atlas of human cell types.