Hey! I’m trying to use a trained TensorFlowLite model in an iOS project, but I’m having some trouble figuring out how to use a TFLite model that takes several inputs in an iOS project.
In fact, this Android project accomplishes exactly what I’d like to do with my TFLite model. However, I see that the API for TensorFlowLiteSwift is a bit more limited than the Android counterpart. Do you all have any advice or any sample projects that successfully use a TFLite model for video classification, or really any data over time on iOS.
Right now, my idea is to just iterate through a window of frames manually feed the interpreter the required inputs at each index. This seems really hacky and prone to breaking, and I’m not even sure it will work as I don’t know if the indices of the input tensors to the interpreter are static or changing, and there is no way (that I know of) to access the signature of each input. But I’ll give it a try until I can think of or find something better.