Skip to content

Latest commit

 

History

History
92 lines (65 loc) · 4.75 KB

README.md

File metadata and controls

92 lines (65 loc) · 4.75 KB
Raw

WildlifeMapper: Aerial Image Analysis for Multi-Species Detection and Identification

WildlifeMapper (WM) is a state-of-the-art model for detecting, locating, and identifying multiple animal species in aerial imagery. It introduces novel modules to enhance localization and identification accuracy, with a verified dataset of 11k images and 28k annotations. This repository contains code for WildlifeMapper, scripts to download and tool to visualize dataset (BisQue).

WildlifeMapper: Aerial Image Analysis for Multi-Species Detection and Identification

Satish Kumar*, Bowen Zhang, .. , Jared A. Stabach, Lacey Hughey, .. , B S Manjunath.

Official repository of our CVPR 2024 paper.

This repository includes:

  • Source code of WildlifeMapper.
  • Pre-trained weights for the bounding box detector.
  • Scripts to download Mara-Wildlife dataset (Approvals under review)
  • Online tool to visualize Mara-Wildlife dataset (BisQue)
  • Code for custom data preparation for training/testing

supported versions Library GitHub license

The repository follows the structure of paper, making it easy to follow and use/extend the work. If this research is helpful to you, please consider citing our paper (bibtex below)

Citing

If this research is helpful to you, please consider citing our paper:

@inproceedings{kumar2024wildlifemapper,
  title={WildlifeMapper: Aerial Image Analysis for Multi-Species Detection and Identification},
  author={Kumar, Satish and Zhang, Bowen and Gudavalli, Chandrakanth and Levenson, Connor and Hughey, Lacey and Stabach, Jared A and Amoke, Irene and Ojwang, Gordon and Mukeka, Joseph and Mwiu, Stephen and others},
  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
  pages={12594--12604},
  year={2024}
}

Usage

Requirements

  • Linux or macOS with Python >= 3.7
  • Pytorch >= 1.7.0
  • CUDA >= 10.0
  • cudNN (compatible with CUDA)

Installation

  1. Clone the repository
  2. Install dependencies
pip install -r requirements.txt

Dataset

See here for an overview of the datastet. The sample dataset can be downloaded here. By downloading the datasets you agree that you have read and accepted the terms of the SA-1B Dataset Research License.

We save masks per image as a json file. It can be loaded as a dictionary in python in the below format.

{
    "image"                 : image_info,
    "annotations"           : [annotation],
}

image_info {
    "image_id"              : int,              # Image id
    "width"                 : int,              # Image width
    "height"                : int,              # Image height
    "file_name"             : str,              # Image filename
}

annotation {
    "id"                    : int,              # Annotation id
    "bbox"                  : [x, y, w, h],     # The box around the mask, in XYWH format
    "predicted_iou"         : float,            # The model's own prediction of the mask's quality
    "stability_score"       : float,            # A measure of the mask's quality
}

License

MethaneMapper is released under the UCSB license. Please see the LICENSE file for more information.

Contributors

The WildlifeMapper project was made possible with the help of many contributors for all over the world: Satish Kumar, Bowen Zhang, Chandrakanth Gudavalli, Connor Levenson, Lacey Hughey, Jared A. Stabach, Irene Amoke, Gordon Ojwang’, Joseph Mukeka, Stephen Mwiu, Joseph Ogutu, Howard Frederick, B.S. Manjunath