Fig. 1: The CAR platform and its neuroscientific applications.

a, The framework of the CAR platform. CAR has a cloud-based architecture and supports diverse types of clients, including workstations, virtual reality (VR) tools, game consoles and mobile apps. CAR is built upon a comprehensive collaboration mechanism, provides management (mgmt.) for data, tasks and users and is boosted by AI capabilities. The system is focused on mouse and human brain data. b, Left, typical morphometric tasks performed using CAR at different reconstruction levels from centimeter to micrometer scales, for example, brain region location tagging, complete neuron reconstruction (WS–VR, neurites annotated by workstation (WS) and validated by VR; WS–WS, VR–WS, VR–VR, similar meanings; check mode, neurites that are reconstructed yet validated), neurite tracing, soma annotation, bouton detection and proofreading. Right, example CAR clients are showcased, including CAR-VR, CAR-Mobile, CAR-Game (also called BrainKiller, unpublished work) and CAR-WS. c, The performance of the CAR server under concurrent scenarios. In CAR, the annotation operations were synchronized among the server and the users using network messages. We modeled situations where many collaborating users (ranging from ten to 100) were sending a burst number of messages. The heatmap shows the average processing time at the CAR server for each message. The y axis indicates the number of messages sent per user, while the x axis represents the number of users engaged in concurrent tasks.