To address the critical scarcity of large-scale, publicly available sonar datasets, we introduce the SonarCloud Dataset for underwater perception tasks. SonarCloud is a comprehensive synthetic dataset generated to accelerate research in underwater perception. The dataset consists of Forward-Looking Sonar (FLS) and depth imagery of 19 distinct objects, along with ground-truth 3D point clouds for each. This dataset is designed to be a valuable tool for the research community, enabling the development and testing of robust underwater perception technologies. Models trained on this data have been shown to successfully detect objects in real-world sonar images and reconstruct their 3D shapes.