src.inference

inference.py

This script performs inference on new time-series data using a trained encoder and a logistic regression classifier. It accepts input files in .hdf5, .csv, or .xlsx format, segments the data, applies the scaler used during training, extracts latent representations with a Keras encoder, and predicts whether each segment corresponds to a ‘walk’ or ‘no_walk’ class using a trained scikit-learn classifier.

The output is a CSV file containing the segment ID, estimated start time, prediction, and mean of the latent vector.

Example

python inference.py –input new_data.hdf5 –encoder encoder_transformer.keras –classifier latent_classifier.pkl –scaler standard_scaler.pkl –output predictions.csv

Functions

load_data(file_path)	Loads input data from .hdf5, .csv, or .xlsx files.
run_inference(input_path, encoder_path, ...)	Runs inference on input data using an encoder and a logistic classifier.
segment_data(data[, segment_len, stride])	Segments time-series data into overlapping windows.