Finescale eddies (FEs) with spatial scales smaller than 50 km play important roles in vertical material transport and energy cascades in the global ocean. However, due to the lack of high-resolution observations, seasonality of FEs is poorly understood. Although the sea surface heights (SSH with resolution of 2 km) measured by the Surface Water and Ocean Topography (SWOT) satellite mission are capable to detect FEs, it is difficult to clearly extract FEs from the SSH which also include internal waves with spatial scales similar to those of FEs. Here, we develop a FE identification method (FEI-Net) by integrating deep learning algorithm, SWOT data and SSH induced by coherent internal tides (main component of internal waves). Verification with in situ observations shows that this method can effectively extract FEs from SWOT data. We constructed a global FE dataset by applying FEI-Net to SWOT data between 2023-2025. The dataset shows that FEs are active in the western boundary current and the Antarctic Circumpolar Current region, and FEs in winter are significantly stronger than those in summer except the equatorial region. It is found that winter enhancement of FEs is mainly modulated by winter-enhanced mixed layer depth. This study revealed the seasonality of FEs from an observational perspective.