pattern-discovery-kit 0.2.0

Pattern Discovery Kit

PATteRn dIsCovery Kit (PATRICK)

Welcome to the Pattern Discovery Kit!

This library provides tools to discover, detect, and track meaningful patterns in physical signals. These signals can be of various forms:

1. Time series,

2. Images,

3. Movies,

4. Any other type of n-dimensional data.

Installation

You can install patrick from the source code by doing the following:

git clone https://github.com/mansour-b/patrick.git
cd patrick
pip install .

Quickstart

Here is an example to briefly present the API:

import numpy as np

from patrick.core import Box, Frame, Model, Movie, Tracker
from patrick.display import plot_frame

# Define the dimensions of the problem
frame_width = 5
frame_height = 5
movie_length = 3
gif_frames_per_second = 2

# Define a concrete model class, just for the example
class DumbModel(Model):
    def predict(self, frame: Frame) -> Frame:
        frame_id = int(frame.name)
        frame.annotations.append(
            Box(label="noise_in_a_square", x=frame_id, y=frame_id, width=1, height=1)
        )
        return frame

model = DumbModel()

# Our data for this short tutorial
frames = [
    Frame(
        name=str(10 * i),
        width=frame_width,
        height=frame_height,
        annotations=[],
        image_array=np.random.random(frame_height, frame_width),
    )
    for i in range(movie_length)
]

movie = Movie(name="some_noise", frames=frames, tracks=[])

# Run the detection model on individual frames
analysed_frames = [model(frame) for frame in movie.frames]
analysed_movie = Movie(
    name="some_noise_with_boxes", frames=analysed_frames, tracks=[]
)

# TBD: run tracker on detections
analysed_movie = tracker.make_tracks(analysed_movie)
analysed_movie.name = "some_noise_with_boxes_and_tracks"

# Plot individual frames with detections
for frame in analysed_movie:
    plot_frame(frame)

# TBD: make a GIF to show the tracks
export_to_gif(analysed_movie, fps=gif_frames_per_seconds)