Aerial Imagery dataset for fire detection: classification and segmentation (Unmanned Aerial Vehicle (UAV))

Overview

Aerial Imagery dataset for fire detection: classification and segmentation using Unmanned Aerial Vehicle (UAV)

Title

FLAME (Fire Luminosity Airborne-based Machine learning Evaluation) Dataset
Alt Text

Paper

You can find the article related to this code here at Elsevier or
You can find the preprint from the Arxiv website.

Dataset

  • The dataset is uploaded on IEEE dataport. You can find the dataset here at IEEE Dataport or DOI. IEEE account is free, so you can create an account and access the dataset files without any payment or subscription.

  • This table below shows all available data for the dataset.

  • This project uses items 7, 8, 9, and 10 from the dataset. Items 7 and 8 are being used for the "Fire_vs_NoFire" image classification. Items 9 and 10 are for the fire segmentation.

  • If you clone this repository on your local drive, please download item 7 from the dataset and unzip in directory /frames/Training/... for the Training phase of the "Fire_vs_NoFire" image classification. The direcotry looks like this:

Repository/frames/Training
                    ├── Fire/*.jpg
                    ├── No_Fire/*.jpg
  • For testing your trained model, please use item 8 and unzip it in direcotry /frame/Test/... . The direcotry looks like this:
Repository/frames/Test
                    ├── Fire/*.jpg
                    ├── No_Fire/*.jpg
  • Items 9 and 10 should be unzipped in these directories frames/Segmentation/Data/Image/... and frames/Segmentation/Data/Masks/... accordingly. The direcotry looks like this:
Repository/frames/Segmentation/Data
                                ├── Images/*.jpg
                                ├── Masks/*.png
  • Please remove other README files from those directories and make sure that only images are there.

Model

  • The binary fire classifcation model of this project is based on the Xception Network:

Alt text

  • The fire segmentation model of this project is based on the U-NET:

Alt text

Sample

  • A short sample video of the dataset is available on YouTube: Alt text

Requirements

  • os
  • re
  • cv2
  • copy
  • tqdm
  • scipy
  • pickle
  • numpy
  • random
  • itertools
  • Keras 2.4.0
  • scikit-image
  • Tensorflow 2.3.0
  • matplotlib.pyplot

Code

This code is run and tested on Python 3.6 on linux (Ubuntu 18.04) machine with no issues. There is a config.py file in this directoy which shows all the configuration parameters such as Mode, image target size, Epochs, batch size, train_validation ratio, etc. All dependency files are available in the root directory of this repository.

  • To run the training phase for the "Fire_vs_NoFire" image classification, change the mode value to 'Training' in the config.py file. Like This
Mode = 'Training'

Make sure that you have copied and unzipped the data in correct direcotry.

  • To run the test phase for the "Fire_vs_NoFire" image classification, change the mode value to 'Classification' in the config.py file. Change This
Mode = 'Classification'

Make sure that you have copied and unzipped the data in correct direcotry.

  • To run the test phase for the Fire segmentation, change the mode value to 'Classification' in the config.py file. Change This
Mode = 'Segmentation'

Make sure that you have copied and unzipped the data in correct direcotry.

Then after setting your parameters, just run the main.py file.

python main.py

Results

  • Fire classification accuracy:

Alt text

  • Fire classification Confusion Matrix:

  • Fire segmentation metrics and evaluation:

Alt text

  • Comparison between generated masks and grount truth mask:

Alt text

  • Federated Learning sample
    To consider future challenges, we defined a new sample of federated learning on a local node (NVidia Jetson Nano, 4GB RAM). Jetson Nano is available in two versions: 1) 4GB RAM developer kit, and 2) 2GB RAM developer kit. In this Implementation, the 4GB version is used with the technical specifications of a 128-core Maxwell GPU, a Quad-core ARM A57 @ 1.43 GHz CPU, 4GB LPDDR4 RAM, and a 32GB microSD storage. To test Jetson Nano for the federated learning, items (9) and (10) from Dataset are used for the fire segmentation. Since Jetson Nano has limited RAM, we assumed that each drone has access to a portion of the FLAME dataset. Only 500 fire images and masks are considered for the training and validation phase on the drone. As we aimed at learning a model on a smaller subset of the FLAME dataset and inferring that model, the default Tensorflow version is used here. Also, the image and mask dimension for each input is reduced to 128 x 128 x 3 rather than 512 x 512 x 3. To save more memory on the RAM, all peripherals were turned off and only WiFi was working at that time for the Secure Shell (SSH) connection. The setup of this node is:

Citation

If you find it useful, please cite our paper as follows:

@article{shamsoshoara2021aerial,
  title={Aerial Imagery Pile burn detection using Deep Learning: the FLAME dataset},
  author={Shamsoshoara, Alireza and Afghah, Fatemeh and Razi, Abolfazl and Zheng, Liming and Ful{\'e}, Peter Z and Blasch, Erik},
  journal={Computer Networks},
  pages={108001},
  year={2021},
  publisher={Elsevier}
}

Other related repositories and articles

License

For academtic and non-commercial usage

Owner
Ph.D. in Informatics and Computing from Northern Arizona University, M.Sc. in Informatics, M.Sc, in Electrical Engineering, B.Sc. in Electrical Engineering
Data and extra materials for the food safety publications classifier

Data and extra materials for the food safety publications classifier The subdirectories contain detailed descriptions of their contents in the README.

1 Jan 20, 2022
banditml is a lightweight contextual bandit & reinforcement learning library designed to be used in production Python services.

banditml is a lightweight contextual bandit & reinforcement learning library designed to be used in production Python services. This library is developed by Bandit ML and ex-authors of Facebook's app

Bandit ML 51 Dec 22, 2022
ALBERT: A Lite BERT for Self-supervised Learning of Language Representations

ALBERT ***************New March 28, 2020 *************** Add a colab tutorial to run fine-tuning for GLUE datasets. ***************New January 7, 2020

Google Research 3k Jan 01, 2023
RM Operation can equivalently convert ResNet to VGG, which is better for pruning; and can help RepVGG perform better when the depth is large.

RM Operation can equivalently convert ResNet to VGG, which is better for pruning; and can help RepVGG perform better when the depth is large.

184 Jan 04, 2023
Unofficial PyTorch implementation of TokenLearner by Google AI

tokenlearner-pytorch Unofficial PyTorch implementation of TokenLearner by Ryoo et al. from Google AI (abs, pdf) Installation You can install TokenLear

Rishabh Anand 46 Dec 20, 2022
Numerical-computing-is-fun - Learning numerical computing with notebooks for all ages.

As much as this series is to educate aspiring computer programmers and data scientists of all ages and all backgrounds, it is also a reminder to mysel

EKA foundation 758 Dec 25, 2022
CaLiGraph Ontology as a Challenge for Semantic Reasoners ([email protected]'21)

CaLiGraph for Semantic Reasoning Evaluation Challenge This repository contains code and data to use CaLiGraph as a benchmark dataset in the Semantic R

Nico Heist 0 Jun 08, 2022
Learnable Multi-level Frequency Decomposition and Hierarchical Attention Mechanism for Generalized Face Presentation Attack Detection

LMFD-PAD Note This is the official repository of the paper: LMFD-PAD: Learnable Multi-level Frequency Decomposition and Hierarchical Attention Mechani

28 Dec 02, 2022
TF2 implementation of knowledge distillation using the "function matching" hypothesis from the paper Knowledge distillation: A good teacher is patient and consistent by Beyer et al.

FunMatch-Distillation TF2 implementation of knowledge distillation using the "function matching" hypothesis from the paper Knowledge distillation: A g

Sayak Paul 67 Dec 20, 2022
Reinforcement Learning for Automated Trading

Reinforcement Learning for Automated Trading This thesis has been realized for the obtention of the Master's in Mathematical Engineering at the Polite

Pierpaolo Necchi 80 Jun 19, 2022
An open source bike computer based on Raspberry Pi Zero (W, WH) with GPS and ANT+. Including offline map and navigation.

Pi Zero Bikecomputer An open-source bike computer based on Raspberry Pi Zero (W, WH) with GPS and ANT+ https://github.com/hishizuka/pizero_bikecompute

hishizuka 264 Jan 02, 2023
Pytorch implementation of Compressive Transformers, from Deepmind

Compressive Transformer in Pytorch Pytorch implementation of Compressive Transformers, a variant of Transformer-XL with compressed memory for long-ran

Phil Wang 118 Dec 01, 2022
PyTorch3D is FAIR's library of reusable components for deep learning with 3D data

Introduction PyTorch3D provides efficient, reusable components for 3D Computer Vision research with PyTorch. Key features include: Data structure for

Facebook Research 6.8k Jan 01, 2023
Data augmentation for NLP, accepted at EMNLP 2021 Findings

AEDA: An Easier Data Augmentation Technique for Text Classification This is the code for the EMNLP 2021 paper AEDA: An Easier Data Augmentation Techni

Akbar Karimi 81 Dec 09, 2022
A series of Jupyter notebooks with Chinese comment that walk you through the fundamentals of Machine Learning and Deep Learning in python using Scikit-Learn and TensorFlow.

Hands-on-Machine-Learning 目的 这份笔记旨在帮助中文学习者以一种较快较系统的方式入门机器学习, 是在学习Hands-on Machine Learning with Scikit-Learn and TensorFlow这本书的 时候做的个人笔记: 此项目的可取之处 原书的

Baymax 1.5k Dec 21, 2022
HyperSeg: Patch-wise Hypernetwork for Real-time Semantic Segmentation Official PyTorch Implementation

: We present a novel, real-time, semantic segmentation network in which the encoder both encodes and generates the parameters (weights) of the decoder. Furthermore, to allow maximal adaptivity, the w

Yuval Nirkin 182 Dec 14, 2022
SpeechNAS Better Trade off between Latency and Accuracy for Large Scale Speaker Verification

SpeechNAS Better Trade off between Latency and Accuracy for Large Scale Speaker Verification

Wentao Zhu 24 May 20, 2022
AFLFast (extends AFL with Power Schedules)

AFLFast Power schedules implemented by Marcel Böhme [email protected]

Marcel Böhme 380 Jan 03, 2023
ChatBot-Pytorch - A GPT-2 ChatBot implemented using Pytorch and Huggingface-transformers

ChatBot-Pytorch A GPT-2 ChatBot implemented using Pytorch and Huggingface-transf

ParZival 42 Dec 09, 2022
Transparent Transformer Segmentation

Transparent Transformer Segmentation Introduction This repository contains the data and code for IJCAI 2021 paper Segmenting transparent object in the

谢恩泽 140 Jan 02, 2023