DeepPeak 0.0.8

Peak detection, trace analysis, and dilution-series tooling for 1D signals.

Meta
Testing
PyPI
Anaconda

DeepPeak

DeepPeak is a Python package for detecting and localizing peaks in 1D signals using deep learning. Designed for researchers and engineers, it simplifies the process of training and deploying neural networks for peak detection.

Key Features

• Deep Learning-based Peak Detection: Leverages Keras and TensorFlow for state-of-the-art performance.

• Gaussian Peak Handling: Built-in support for detecting Gaussian-shaped peaks.

• Custom Signal Support: Easily adaptable to various types of 1D signals.

• Easy-to-Use API: Train and predict with minimal setup.

Analysis Quickstart

For dilution-series analysis, prefer the namespaced DilutionSeries API and run the standard and CNN detectors explicitly:

from DeepPeak.analysis import DilutionSeries, HeightPeakTrigger, SigmaPeakTrigger

series = DilutionSeries(
    folder="path/to/traces",
    wavenet=wavenet,
    initial_concentration=1.0,
    nrows=100_000,
)

standard = series.run_standard(
    std_trigger=SigmaPeakTrigger(sigma=5.0, hysteresis=4.0),
)

cnn = series.run_cnn(
    cnn_trigger=HeightPeakTrigger(height=0.05, hysteresis=0.03),
    cnn_amplitude_sigma_samples=27,
)

series.plot.standard_detection(index=0)
series.plot.wavenet_detection(index=0)

series.poisson.plot.expected_histogram(
    index=0,
    base_index=0,
    detector="standard",
    x_axis="time",
)

series.amplitude.plot.histogram(index=0, detector="standard")
series.width.plot.histogram(index=0, detector="standard", x_axis="time")

If you want the old all-in-one execution, series.run() still exists, but it now requires both the standard and CNN detector configurations to be present.

The older PeakCountSeries name remains available as a backward-compatible alias, but new notebook code should prefer DilutionSeries.

Contact

For questions or contributions, contact [e-mail].