Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Streamlined Hybrid Annotation Framework using Scalable Codestream for Bandwidth-Restricted UAV Object Detection

Karim El Khoury, Tiffanie Godelaine, Simon Delvaux, Sebastien Lugan, Benoit Macq

TL;DR

Emergency UAV workflows are constrained by limited bandwidth, which delays critical decisions. The paper introduces a bandwidth-aware streamlined hybrid annotation framework that leverages a JPEG 2000 scalable codestream to run a fine-tuned DL detector on low-resolution data and selectively fetch high-resolution tiles for human annotation in regions of interest, achieving up to a $34$-fold reduction in decision time. Across nine emergency scenarios and varying data rates, the approach maintains competitive recall while dramatically accelerating response, illustrating the practicality of on-board, multi-resolution transmission and selective high-detail annotation. This work enables faster, bandwidth-adaptive UAV object detection and points to future enhancements via few-shot and active learning for continued performance gain in dynamic emergency contexts.

Abstract

Emergency response missions depend on the fast relay of visual information, a task to which unmanned aerial vehicles are well adapted. However, the effective use of unmanned aerial vehicles is often compromised by bandwidth limitations that impede fast data transmission, thereby delaying the quick decision-making necessary in emergency situations. To address these challenges, this paper presents a streamlined hybrid annotation framework that utilizes the JPEG 2000 compression algorithm to facilitate object detection under limited bandwidth. The proposed framework employs a fine-tuned deep learning network for initial image annotation at lower resolutions and uses JPEG 2000's scalable codestream to selectively enhance the image resolution in critical areas that require human expert annotation. We show that our proposed hybrid framework reduces the response time by a factor of 34 in emergency situations compared to a baseline approach.

Streamlined Hybrid Annotation Framework using Scalable Codestream for Bandwidth-Restricted UAV Object Detection

TL;DR

Emergency UAV workflows are constrained by limited bandwidth, which delays critical decisions. The paper introduces a bandwidth-aware streamlined hybrid annotation framework that leverages a JPEG 2000 scalable codestream to run a fine-tuned DL detector on low-resolution data and selectively fetch high-resolution tiles for human annotation in regions of interest, achieving up to a -fold reduction in decision time. Across nine emergency scenarios and varying data rates, the approach maintains competitive recall while dramatically accelerating response, illustrating the practicality of on-board, multi-resolution transmission and selective high-detail annotation. This work enables faster, bandwidth-adaptive UAV object detection and points to future enhancements via few-shot and active learning for continued performance gain in dynamic emergency contexts.

Abstract

Emergency response missions depend on the fast relay of visual information, a task to which unmanned aerial vehicles are well adapted. However, the effective use of unmanned aerial vehicles is often compromised by bandwidth limitations that impede fast data transmission, thereby delaying the quick decision-making necessary in emergency situations. To address these challenges, this paper presents a streamlined hybrid annotation framework that utilizes the JPEG 2000 compression algorithm to facilitate object detection under limited bandwidth. The proposed framework employs a fine-tuned deep learning network for initial image annotation at lower resolutions and uses JPEG 2000's scalable codestream to selectively enhance the image resolution in critical areas that require human expert annotation. We show that our proposed hybrid framework reduces the response time by a factor of 34 in emergency situations compared to a baseline approach.
Paper Structure (16 sections, 7 equations, 3 figures, 1 table, 5 algorithms)

This paper contains 16 sections, 7 equations, 3 figures, 1 table, 5 algorithms.

Table of Contents

  1. Introduction
  2. Related Works
  3. Related works merging (I) and (IV):
  4. Related works merging (I) and (II):
  5. Related works merging (I), (II) and (IV):
  6. Related works merging (II), (III) and (IV):
  7. Proposed Approach
  8. Baseline hybrid annotation framework
  9. Proposed streamlined hybrid annotation framework
  10. Experimental setup
  11. Materials
  12. Evaluation metrics
  13. Experiments
  14. Results and discussion
  15. Conclusion
  16. ...and 1 more sections

Figures (3)

  • Figure 1: Simplified block diagram of the baseline hybrid annotation framework.
  • Figure 2: Simplified block diagram of the proposed streamlined hybrid annotation framework.
  • Figure 3: Recall versus time plot for the baseline framework and the proposed framework for high-resolution image size of 60 MP under a data rate limit of (a) 176 kbps (b) 88 kbps and (c) 22 kbps, with a maximum human annotator time of 5 minutes and considering emergency scenarios with a transmission time limit of: 3, 10, and 30 minutes.