There's no go-to formula for any ML project. A huge part of improving machine learning systems is experimenting until you find a solution that works well for your use case. However, an overwhelming amount of tools and model architectures are available for computer vision tasks.