Unifying Remote Sensing Image Retrieval and Classification with Robust Fine-tuning

Abstract

Advances in high resolution remote sensing image analysis are currently hampered by the difficulty of gathering enough annotated data for training deep learning methods, giving rise to a variety of small datasets and associated dataset-specific methods. Moreover, typical tasks such as classification and retrieval lack a systematic evaluation on standard benchmarks and training datasets, which make it hard to identify durable and generalizable scientific contributions. We aim at unifying remote sensing image retrieval and classification with a new large-scale training and testing dataset, SF300 1 , including both vertical and oblique aerial images and made available to the research community, and an associated finetuning method. We additionally propose a new adversarial fine-tuning method for global descriptors. We show that our framework systematically achieves a boost of retrieval and classification performance on nine different datasets compared to an ImageNet pretrained baseline, with currently no other method to compare to.

Connecting Images through Time and Sources: Introducing Low-data, Heterogeneous Instance Retrieval

“…Table III shows the performance of GeM on ALEGORIA depending on the training dataset (with our implementation and training protocol). We pick three datasets corresponding to three types of situations encountered in ALEGORIA: GoogleLandmarks [9] (1.4M images, 81k classes, landmark images taken mostly from the ground), SF300 [12] (308k images, 27k classes, aerial vertical and oblique photography) and University1652 [13] (50k images, 701 classes, landmark images in a cross-view setup). We additionaly conduct experiments with Imagenet and SfM-120k (120k images, 55 classes) for reference.…”