I plan to use federated learning for an object detection algorithm I already developed for detecting weeds. As I research, I see federated tensorflow examples on Image classification. Like the following link: https://www.tensorflow.org/federated/tutorials/federated_learning_for_image_classification
My question is can we use federated learning and federated tensorflow for object detection algorithms? If yes, would you please provide me with some links and examples?
TFF can be used with arbitrary TensorFlow; generally, all the symbols in the tff.learning namespace are really convenience wrappers for common use cases of the lower-level API, the federated core.
I'm not personally aware of any 'FL for object detection' research, but there appears to have been some projects in this direction previously. I'm no expert in object detection, but generally it seems to me that there may be some interesting questions around e.g. labels in the federated setting for this kind of application.
That is, combining the two thrusts above: I think you are interested in a research area that has gotten some thought but not a tremendous amount yet, and therefore you are likely to be writing your own federated algorithms from scratch. TFF may certainly be a useful tool here; see for example this standalone implementation of FedAvg for another perspective on writing a custom algorithm in TFF.
