Learning to Interpret Satellite Images in Global Scale Using Wikipedia

Abstract

Despite recent progress in computer vision, finegrained interpretation of satellite images remains challenging because of a lack of labeled training data. To overcome this limitation, we construct a novel dataset called WikiSatNet by pairing georeferenced Wikipedia articles with satellite imagery of their corresponding locations. We then propose two strategies to learn representations of satellite images by predicting properties of the corresponding articles from the images. Leveraging this new multi-modal dataset, we can drastically reduce the quantity of human-annotated labels and time required for downstream tasks. On the recently released fMoW dataset, our pre-training strategies can boost the performance of a model pre-trained on ImageNet by up to 4:5% in F1 score.

Accepted to IJCAI 2019

References in corpus (6)

• Adam: A Method for Stochastic Optimization
• Efficient Estimation of Word Representations in Vector Space
• Semi-Supervised Learning with Deep Generative Models
• Snorkel: Rapid Training Data Creation with Weak Supervision
• Learning to Interpret Satellite Images Using Wikipedia
• Adversarial Constraint Learning for Structured Prediction

Cited by in corpus (6)

• Efficient Poverty Mapping using Deep Reinforcement Learning
• Hard-Attention for Scalable Image Classification
• RarePlanes: Synthetic Data Takes Flight
• Effective Data Fusion with Generalized Vegetation Index: Evidence from Land Cover Segmentation in Agriculture
• Farmland Parcel Delineation Using Spatio-temporal Convolutional Networks
• Predicting Poverty Level from Satellite Imagery using Deep Neural Networks