Track Boosting and Synthetic Data Aided Drone Detection

Akyon, Fatih Çağatay; Eryuksel, Ogulcan; Ozfuttu, Kamil Anil; Altinuc, Sinan Onur

doi:10.1109/avss52988.2021.9663759

Cited by 25 publications

(24 citation statements)

References 12 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This would normally result in the reduction of most animals to a few pixels as the Faster R‐CNN needs to downscale the pictures for processing. Therefore, we used slicing aided hyper inference (SAHI) to ensure good predictions regardless of size (Akyon et al, 2021, 2022). Within the SAHI framework, images are segmented and predictions are made on the individual slices, removing the need to load extremely large images.…”

Section: Methodsmentioning

confidence: 99%

CollembolAI, a macrophotography and computer vision workflow to digitize and characterize samples of soil invertebrate communities preserved in fluid

et al. 2022

View full text Add to dashboard Cite

Soils are inhabited by communities of tiny invertebrates that participate in the essential functions of soils. Characterizing those communities in terms of species diversity and species abundance is part of investigating soil functions and response to perturbation. Dozens to hundreds of specimens can be extracted from a sample that need to be sorted, counted and identified. It involves an enormous amount of time, straining the workflow of soil zoologists. Deep learning‐based computer vision approaches have become increasingly popular to monitor biodiversity, as they can be applied to detect, count and classify organisms. In this work, we present CollembolAI, an open‐source prototype for a computer vision workflow. It includes a hardware system for acquiring high‐resolution pictures of soil species samples in fluid and a deep learning‐based application (Faster R‐CNN with Slicing Aided Hyper Inference) to train and evaluate models for the detection and classification of animals on those pictures. We evaluated the workflow using a mix of specimens belonging to 12 species of springtail and mite, picked from our taxonomic collection. Specimens were photographed multiple times under various angle of views. The model was trained using 5671 views on specimens on 30 images and tested on 442 views from new specimens on six images. CollembolAI is affordable, simple to build and allows the rapid digitization of mesofauna samples. Our deep learning model achieved a Precision of 0.940, Recall of 0.918 and a mAP@0.5 (Pascal VOC) of 0.868. The model showed a lower Recall for one species, but was performant on all others. Our prototype offers an operational workflow for the creation of soil fauna picture datasets needed to develop efficient deep learning‐based classifiers. The applications are numerous, for example, collection digitization, soil biodiversity analysis and monitoring, or automatic assessment of mesofauna‐based bioindicators. Computer vision is a rapidly emerging tool to handle efficiently the complexity of biodiversity and is already successfully used for plants and vertebrates recognition of pictures. It is on the way to become a major asset for dealing with invertebrate diversity. The code and setup instructions are available on Github.

show abstract

Section: Methodsmentioning

confidence: 99%

CollembolAI, a macrophotography and computer vision workflow to digitize and characterize samples of soil invertebrate communities preserved in fluid

et al. 2022

View full text Add to dashboard Cite

show abstract

“…An initial approach for drone detection is to adopt state-of-the-art ConvNets, including SSD 19 , Faster R-CNN 42 , Yolov2 43 , Yolov4 20 and more recently Yolov5 44,45 . Because these models are large and complex, synthetic data are required and the real-time constraint is removed.…”

Section: Drone Detectionmentioning

confidence: 99%

Small Flying Object Detection and Tracking in Digital Airport Tower through Spatial-Temporal ConvNets

Thai

Alam

Lilith

et al. 2023

Preprint

View full text Add to dashboard Cite

Digital towers using high-resolution cameras that cover a 360-degree view of airports have recently been applied as a solution for some airports by replacing conventional towers. Although many computer vision systems have been developed as tools to assist tower controllers, small flying object detection remains challenging due to their small dimensions and unpredictable trajectories. This paper proposes a novel computer vision framework to detect, track and recognize small flying objects, namely aircraft and drones, in an airport environment. The framework creates a new Convolutional Neural Network which adapts to the unique characteristics of small flying objects. It also exploits the spatial-temporal information, as well as post-processing, to improve the performance. The proposed framework is validated on an airport dataset and Drone-vs-Bird public dataset. The results show that the framework can not only perform object detection in real-time, but also surpass the performance of state-of-the-art models in both datasets by a large margin.

show abstract

“…• Track Boosting and Synthetic Data Aided Drone Detection [13] -Authors use two-stage solution. First generating synthetic data to extend the data set, then apply object detection model.…”

Section: Literature Reviewmentioning

confidence: 99%

Bird/Drone Detection and Classification using Classical and Deep Learning Methods

Zitar

Kassab

Fallah

et al. 2023

Preprint

View full text Add to dashboard Cite

Machine learning techniques have always been a strong candidate for solving complex recognition problems. Drone/Bird detection and classification is one of the most challenging tasks in recent years. Both drones and birds come in different sizes, velocities, and behaviors. The lack of bird images and videos is tackled in this work. Deep learning, classical machine learning techniques such as Support Vector Machines (SVM) and Random Forest (RF) in addition to shallow neural network (NN) learning methods are used. Combined open-source data sets and labeled bird images data sets are used in training and testing for detection and classification. In particular, several deep learning methods are used in the detection of RGB and IR drone images. They were compared with the new SSD-AdderNet which showed the best results in the detection of IR images while exhibiting the least complexity. The SVM proved to be the best in classification.

show abstract

Track Boosting and Synthetic Data Aided Drone Detection

Cited by 25 publications

References 12 publications

CollembolAI, a macrophotography and computer vision workflow to digitize and characterize samples of soil invertebrate communities preserved in fluid

CollembolAI, a macrophotography and computer vision workflow to digitize and characterize samples of soil invertebrate communities preserved in fluid

Small Flying Object Detection and Tracking in Digital Airport Tower through Spatial-Temporal ConvNets

Bird/Drone Detection and Classification using Classical and Deep Learning Methods

Contact Info

Product

Resources

About