On-Board Small-Scale Object Detection for Unmanned Aerial Vehicles (UAVs)
Object detection is a critical task that becomes difficult when dealing with onboard detection using aerial images and computer vision technique. The main challenges with aerial images are small target sizes, low resolution, occlusion, attitude, and scale variations, which affect the performance of...
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2023-05-01
|
| Series: | Drones |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2504-446X/7/5/310 |
| _version_ | 1797600441229377536 |
|---|---|
| author | Zubair Saeed Muhammad Haroon Yousaf Rehan Ahmed Sergio A. Velastin Serestina Viriri |
| author_facet | Zubair Saeed Muhammad Haroon Yousaf Rehan Ahmed Sergio A. Velastin Serestina Viriri |
| author_sort | Zubair Saeed |
| collection | DOAJ |
| description | Object detection is a critical task that becomes difficult when dealing with onboard detection using aerial images and computer vision technique. The main challenges with aerial images are small target sizes, low resolution, occlusion, attitude, and scale variations, which affect the performance of many object detectors. The accuracy of the detection and the efficiency of the inference are always trade-offs. We modified the architecture of CenterNet and used different CNN-based backbones of ResNet18, ResNet34, ResNet50, ResNet101, ResNet152, Res2Net50, Res2Net101, DLA-34, and hourglass14. A comparison of the modified CenterNet with nine CNN-based backbones is conducted and validated using three challenging datasets, i.e., VisDrone, Stanford Drone dataset (SSD), and AU-AIR. We also implemented well-known off-the-shelf object detectors, i.e., YoloV1 to YoloV7, SSD-MobileNet-V2, and Faster RCNN. The proposed approach and state-of-the-art object detectors are optimized and then implemented on cross-edge platforms, i.e., NVIDIA Jetson Xavier, NVIDIA Jetson Nano, and Neuro Compute Stick 2 (NCS2). A detailed comparison of performance between edge platforms is provided. Our modified CenterNet combination with hourglass as a backbone achieved 91.62%, 75.61%, and 34.82% mAP using the validation sets of AU-AIR, SSD, and VisDrone datasets, respectively. An FPS of 40.02 was achieved using the ResNet18 backbone. We also compared our approach with the latest cutting-edge research and found promising results for both discrete GPU and edge platforms. |
| first_indexed | 2024-03-11T03:48:10Z |
| format | Article |
| id | doaj.art-e66f090a80fd49089cacf2fde1f03f69 |
| institution | Directory Open Access Journal |
| issn | 2504-446X |
| language | English |
| last_indexed | 2024-03-11T03:48:10Z |
| publishDate | 2023-05-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Drones |
| spelling | doaj.art-e66f090a80fd49089cacf2fde1f03f692023-11-18T01:07:08ZengMDPI AGDrones2504-446X2023-05-017531010.3390/drones7050310On-Board Small-Scale Object Detection for Unmanned Aerial Vehicles (UAVs)Zubair Saeed0Muhammad Haroon Yousaf1Rehan Ahmed2Sergio A. Velastin3Serestina Viriri4Swarm Robotics Lab (SRL), National Center of Robotics and Automation (NCRA), University of Engineering and Technology (UET), Taxila 47080, PakistanSwarm Robotics Lab (SRL), National Center of Robotics and Automation (NCRA), University of Engineering and Technology (UET), Taxila 47080, PakistanSchool of Electrical Engineering and Computer Science, National University of Sciences and Technology, Islamabad 24090, PakistanSchool of Electronic Engineering and Computer Science, Queen Mary University of London, London E1 4NS, UKSchool of Mathematics, Statistics and Computer Science, University of KwaZulu-Natal, Durban 4041, South AfricaObject detection is a critical task that becomes difficult when dealing with onboard detection using aerial images and computer vision technique. The main challenges with aerial images are small target sizes, low resolution, occlusion, attitude, and scale variations, which affect the performance of many object detectors. The accuracy of the detection and the efficiency of the inference are always trade-offs. We modified the architecture of CenterNet and used different CNN-based backbones of ResNet18, ResNet34, ResNet50, ResNet101, ResNet152, Res2Net50, Res2Net101, DLA-34, and hourglass14. A comparison of the modified CenterNet with nine CNN-based backbones is conducted and validated using three challenging datasets, i.e., VisDrone, Stanford Drone dataset (SSD), and AU-AIR. We also implemented well-known off-the-shelf object detectors, i.e., YoloV1 to YoloV7, SSD-MobileNet-V2, and Faster RCNN. The proposed approach and state-of-the-art object detectors are optimized and then implemented on cross-edge platforms, i.e., NVIDIA Jetson Xavier, NVIDIA Jetson Nano, and Neuro Compute Stick 2 (NCS2). A detailed comparison of performance between edge platforms is provided. Our modified CenterNet combination with hourglass as a backbone achieved 91.62%, 75.61%, and 34.82% mAP using the validation sets of AU-AIR, SSD, and VisDrone datasets, respectively. An FPS of 40.02 was achieved using the ResNet18 backbone. We also compared our approach with the latest cutting-edge research and found promising results for both discrete GPU and edge platforms.https://www.mdpi.com/2504-446X/7/5/310object detectioncomputer visionmodified CenterNetVisDroneSSDAU-AIR |
| spellingShingle | Zubair Saeed Muhammad Haroon Yousaf Rehan Ahmed Sergio A. Velastin Serestina Viriri On-Board Small-Scale Object Detection for Unmanned Aerial Vehicles (UAVs) Drones object detection computer vision modified CenterNet VisDrone SSD AU-AIR |
| title | On-Board Small-Scale Object Detection for Unmanned Aerial Vehicles (UAVs) |
| title_full | On-Board Small-Scale Object Detection for Unmanned Aerial Vehicles (UAVs) |
| title_fullStr | On-Board Small-Scale Object Detection for Unmanned Aerial Vehicles (UAVs) |
| title_full_unstemmed | On-Board Small-Scale Object Detection for Unmanned Aerial Vehicles (UAVs) |
| title_short | On-Board Small-Scale Object Detection for Unmanned Aerial Vehicles (UAVs) |
| title_sort | on board small scale object detection for unmanned aerial vehicles uavs |
| topic | object detection computer vision modified CenterNet VisDrone SSD AU-AIR |
| url | https://www.mdpi.com/2504-446X/7/5/310 |
| work_keys_str_mv | AT zubairsaeed onboardsmallscaleobjectdetectionforunmannedaerialvehiclesuavs AT muhammadharoonyousaf onboardsmallscaleobjectdetectionforunmannedaerialvehiclesuavs AT rehanahmed onboardsmallscaleobjectdetectionforunmannedaerialvehiclesuavs AT sergioavelastin onboardsmallscaleobjectdetectionforunmannedaerialvehiclesuavs AT serestinaviriri onboardsmallscaleobjectdetectionforunmannedaerialvehiclesuavs |