Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Bootstrapping Rare Object Detection in High-Resolution Satellite Imagery

Akram Zaytar, Caleb Robinson, Gilles Q. Hacheme, Girmaw A. Tadesse, Rahul Dodhia, Juan M. Lavista Ferres, Lacey F. Hughey, Jared A. Stabach, Irene Amoke

TL;DR

A significant enhancement in detection efficiency is demonstrated, achieving a positive sampling rate increase from 2% (random) to 30%.

Abstract

Rare object detection is a fundamental task in applied geospatial machine learning, however is often challenging due to large amounts of high-resolution satellite or aerial imagery and few or no labeled positive samples to start with. This paper addresses the problem of bootstrapping such a rare object detection task assuming there is no labeled data and no spatial prior over the area of interest. We propose novel offline and online cluster-based approaches for sampling patches that are significantly more efficient, in terms of exposing positive samples to a human annotator, than random sampling. We apply our methods for identifying bomas, or small enclosures for herd animals, in the Serengeti Mara region of Kenya and Tanzania. We demonstrate a significant enhancement in detection efficiency, achieving a positive sampling rate increase from 2% (random) to 30%. This advancement enables effective machine learning mapping even with minimal labeling budgets, exemplified by an F1 score on the boma detection task of 0.51 with a budget of 300 total patches.

Bootstrapping Rare Object Detection in High-Resolution Satellite Imagery

TL;DR

A significant enhancement in detection efficiency is demonstrated, achieving a positive sampling rate increase from 2% (random) to 30%.

Abstract

Rare object detection is a fundamental task in applied geospatial machine learning, however is often challenging due to large amounts of high-resolution satellite or aerial imagery and few or no labeled positive samples to start with. This paper addresses the problem of bootstrapping such a rare object detection task assuming there is no labeled data and no spatial prior over the area of interest. We propose novel offline and online cluster-based approaches for sampling patches that are significantly more efficient, in terms of exposing positive samples to a human annotator, than random sampling. We apply our methods for identifying bomas, or small enclosures for herd animals, in the Serengeti Mara region of Kenya and Tanzania. We demonstrate a significant enhancement in detection efficiency, achieving a positive sampling rate increase from 2% (random) to 30%. This advancement enables effective machine learning mapping even with minimal labeling budgets, exemplified by an F1 score on the boma detection task of 0.51 with a budget of 300 total patches.
Paper Structure (11 sections, 1 equation, 2 figures, 1 table)

This paper contains 11 sections, 1 equation, 2 figures, 1 table.

Table of Contents

  1. Introduction
  2. Problem Formulation
  3. Methods
  4. Initializing the sampling surface
  5. Sampling strategies
  6. Case Study: Boma Mapping in the Serengeti Mara
  7. Bootstrapping
  8. Downstream training
  9. Results and Discussion
  10. Conclusion
  11. Unsupervised method for choosing hyperparameters

Figures (2)

  • Figure 1: Bootstrapping rare object detection. Given input imagery, we create a grid of image patches, initialize a sampling surface over the same grid, and sample iteratively from the surface looking for rare object instances until we hit a budget limit. Sampling strategies that surface these rare objects more frequently than random allow for quicker instantiating of model based methods for finding such objects.
  • Figure 2: Silhouette scores of different clustering methods versus the expected number of samples required to find 100 positives with an offline sampling scheme.