It depends on your image resolution and the type of object which you wish to detect.
For example. If I want to detect a car or tree or road crosswalks I want 20 to 30 cm pixel size. On another side, If I want to detect forest patches, land parcels, or settlement clusters I can use 50 cm to 1-meter resolution.
Similarly, If you are working on Land Use and Land Cover you can work 30 cm to 10-meter resolution. Even you can use deep learning on sentinel datasets. Hope I answered your question.