LeOpardLink 0.1.0

A package to generate all possible graphs for wildlife individual identification, and visualize them

LeOpardLink

Welcome to LeOpard Link (LOL)

This package is designed to assemble visual networks of detections of marked animal individuals (e.g., leopards, snow leopards, tigers, etc.) from camera-trapping datasets. We define detections as nodes in the graph, and the relationships of any pair of detections to be edges in the graph. The edges can be certain (two detections are/aren't the same individual), or uncertain (we do not know if the two detections are the same or not, e.g., detection 1 captures the left flank while detection 2 captures the right flank).

While this package was motivated for the use of wildlife camera trap images, it has many applications in other picture-relationship-based projects.

Installation

To install the package, download our repo first. Under the repo directory, use the following command:

pip install -e .

Dependencies

The package requires the following dependencies:

• numpy

• pandas

• networkx

• jaal>=0.1.7

Features

• Network Visualization: Generate an interactive plot of the current network.

• Graph Generation: Generate all possible graphs from an identification matrix with uncertain edges.

• Individual Counts: Calculate the number of individuals, mean, standard deviation, and 95% confidence interval.

Functions

• checkInput: Check if the input adjacency matrix is valid. The identification matrix should only contain 3 numeric values: 1 (two detections are the same individual), 0 (two detections are not the same individual), -1 (uncertain relationship between the two detections) .

• createAdjList: Create an adjacency list from an adjacency matrix. Adjcency list and matrix are two different ways to store graph information.

• adjListToMatrix: Convert an adjacency list to an adjacency matrix.

• sumWeights: Calculate the sum of weights connected to a node.

• checkSymmetric: Check if the graph is symmetric.

• checkTransitivityWeighted: Check if the graph is transitive.

• detect_conflicts: Detect conflicts in the adjacency list.

• strictTransitiveClosure: Compute the strict transitive closure of an adjacency matrix.

• generateGraphsWithTransitivity: Generate all possible graphs with transitivity from an adjacency list.

• GraphProperty: Generate a dataframe of all graphs and their properties, including the number of clusters (connected nodes).

• JaalDataPrepareNode: Prepare node data for Jaal plotting.

• JaalDataPrepareEdge: Prepare edge data for Jaal plotting.

• JaalPlot: Plot the graph using Jaal.

• simulationMatrix: Generate a simulation adjacency matrix.

How does it work?

See the network below for a simulation dataset. 21 nodes represent 21 images from wildlife camera traps. Known matches are represented by edges. Based on this figure, there are 7 individuals.

import numpy as np
from LeOpardLink import matrices

# Example adjacency matrix
example_matrix = matrices.simulationMatrix()

# Check input
matrices.checkInput(example_matrix)

# Create adjacency list
adj_list = matrices.createAdjList(example_matrix)

# Check symmetry
matrices.checkSymmetric(adj_list)

# Check transitivity
is_transitive = matrices.checkTransitivityWeighted(adj_list)

# Generate graphs with transitivity
all_graphs = matrices.generateGraphsWithTransitivity(adj_list)

# Get graph properties
graph_properties = matrices.GraphProperty(all_graphs)

# Plot the graph using Jaal
G = nx.from_numpy_array(allgraphs[0])
node_df = matrices.JaalDataPrepareNode(G)
edge_df = matrices.JaalDataPrepareEdge(G)
matrices.JaalPlot(node_df, edge_df)

Contributing

Contributions are welcome! Please see the issues page for ways you can help.

License

This project is licensed under the MIT License. See the LICENSE file for details.

Authors

Courtney Allen

• PLEASE WRITE YOUR CONTRIBUTIONS HERE

Ziqi Guo

• PLEASE WRITE YOUR CONTRIBUTIONS HERE

Guy Bennevat Haninovich

• PLEASE WRITE YOUR CONTRIBUTIONS HERE

Jiangyue Wang

• Wrote python functions
• Wrote tests
• Wrote documentation
• Set up packaging