Cities are complex systems that are constantly changing. This paper explores the capabilities of using crowdsourced street-level imagery in observing city dynamics. Visual walkability is an example of such an index, where different results may be obtained depending on locational and temporal factors. This paper introduces a new index called Type of Visual Walkability (TVW) to characterize and classify street-level visual walkability in Inner London utilizing Mapillary images. The method is based on panoptic segmentation, where pixel-level segmentation and instance count are used in combination to generate more robust indicators of greenery, openness, crowdedness, and visual pavement. Following this, the TVW at street segment level is calculated and the spatiotemporal dynamics of TVW are explored. The results show significant seasonal variations. Specifically, many greenery-dominated streets become openness-dominated from autumn to winter and pavement-dominated streets become crowdedness-dominated in summer and autumn due to vegetation phenology and human activities. This case study showed that TVW provides a dynamic and explainable perspective in understanding urban design qualities for walkability. It facilitates the connection between assessment of the built environment and spatiotemporal analysis derived from street-level images and will inform urban planners and governments in building a walkable city and further promote active transport.