As groups of coeval stars born from the same molecular cloud, an open cluster (OC) is an ideal laboratory for studying the structure and dynamical evolution of the Milky Way. The release of high-precision Gaia Early Data Release 3 (Gaia EDR3) and modern machine-learning methods offer unprecedented opportunities to identify OCs. In this study, we extended conventional HDBSCAN (e-HDBSCAN) for searching for new OCs in Gaia EDR3. A pipeline was developed based on the parallel computing technique to blindly search for OCs from Gaia EDR3 within Galactic latitudes |b|<25{deg}. As a result, we obtained 3787 star clusters, of which 83 new OCs were reported after cross-match and visual inspection. At the same time, the main star cluster parameters are estimated by color-magnitude diagram fitting. The study significantly increases the sample size and physical parameters of OCs in the catalog of OCs. It shows the incompleteness of the census of OCs across our Galaxy.