Accurately locating multiple objects and individuals within indoor spaces poses a significant challenge. This issue holds immense importance for Internet of Things (IoT) applications. In numerous scenarios, there exists a high density of people and objects requiring precise spatiotemporal tracking. As IoT applications expand beyond industrial settings to areas like healthcare facilities, addressing indoor localization limitations is crucial. This paper presents an innovative indoor localization technique achieving 10 cm resolution in both Line-of-Sight (LoS) and Non-Line-of-Sight (NLoS) configurations. While Received Signal Strength Indication (RSSI) based on WiFi signals has been extensively studied for localization solutions, it is highly susceptible to temporal and spatial variations due to multipath effects. In contrast, the Channel State Information (CSI) transmitted via WiFi hardware provides spatial and temporal data remaining relatively stable amid multipath propagation and interference, making it highly suitable for precise localization design. The proposed solution utilizes open-source hardware. A CSI fingerprint database is generated for 100 locations spaced 10 cm apart in both LOS and NLOS environments. Various machine learning algorithms are employed to generate classifier models. Notably, the Random Forest classifier achieves 93.15% accuracy for LOS and 98.01% for NLOS at 10 cm resolution, representing a seminal achievement. This design can potentially extend to diverse applications leveraging existing WiFi infrastructure, including patient localization in crowded hospital waiting areas, home healthcare, and tracking multiple tools in industrial settings. The key technical details and findings are maintained while using more concise language. Please advise if any part requires further clarification or improvement.