iSAID: A Large-scale Dataset for Instance Segmentation in Aerial Images

By Syed Waqas Zamir, Aditya Arora, Akshita Gupta, Salman Hameed Khan, Guolei Sun, Fahad Shahbaz Khan, Fan Zhu, Ling Shao, Guisong Xia, Xiang Bai. 2019

507 citations recorded by the source corpus.

Existing Earth Vision datasets are either suitable for semantic segmentation or object detection. In this work, we introduce the first benchmark dataset for instance segmentation in aerial imagery that combines instance-level object detection and pixel-level segmentation tasks. In comparison to instance segmentation in natural scenes, aerial images present unique challenges e.g., a huge number of instances per image, large object-scale variations and abundant tiny objects. Our large-scale and densely annotated Instance Segmentation in Aerial Images Dataset (iSAID) comes with 655,451 object instances for 15 categories across 2,806 high-resolution images. Such precise per-pixel annotations for each instance ensure accurate localization that is essential for detailed scene analysis. Compared to existing small-scale aerial image based instance segmentation datasets, iSAID contains 15$\times$ the number of object categories and 5$\times$ the number of instances. We benchmark our dataset using two popular instance segmentation approaches for natural images, namely Mask R-CNN and PANet. In our experiments we show that direct application of off-the-shelf Mask R-CNN and PANet on aerial images provide suboptimal instance segmentation results, thus requiring specialized solutions from the research community. The dataset is publicly available at: this https URL

Subfields: Aerospace Engineering, Computer Vision and Pattern Recognition

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Aerospace Engineering

In Computer Vision and Pattern Recognition

Citations in the corpus

Satlas: A Large-Scale, Multi-Task Dataset for Remote Sensing Image Understanding (2022, methodology): “iSAID [54] involves instance segmentation for 15 categories.”
HRSID: A High-Resolution SAR Images Dataset for Ship Detection and Instance Segmentation (2020, methodology): “So, distinguished from constructing an optical remote sensing dataset for object detection and instance segmentation [54], ships should be accurately and completely annotated when constructing the SAR dataset for ship detection and instance segmentation.”
Object Detection in Aerial Images: A Large-Scale Benchmark and Challenges (2021, methodology): “TAS [26] satellite HBB 1 1 1,319 30 792 2008 SZTAKI-INRIA [7] multi source OBB 1 1 665 9 800 2012 NWPU VHR-10 [42] multi source HBB 10 10 3,651 800 1000 2014 VEDAI [15] satellite OBB 3 9 2,950 1,268 512, 1024 2015 DLR 3k [16] aerial OBB 2 8 14,235 20 5616 2015 UCAS-AOD [17] Google Earth OBB 2 2 14,5”
An Empirical Study of Remote Sensing Pretraining (2022, background): “large-scale segmentation benchmark — iSAID [75], to serve as the test bed of the corresponded subtasks, respectively.”
EarthGPT: A Universal Multimodal Large Language Model for Multisensor Image Comprehension in Remote Sensing Domain (2024, methodology): “Segmentation datasets typically include Vaihingen, Potsdam, iSAID [69], and LoveDA [70].”
SegNeXt: Rethinking Convolutional Attention Design for Semantic Segmentation (2022, methodology): “We evaluate our methods on seven popular datasets, including ImageNet-1K [15], ADE20K [98], Cityscapes [13], Pascal VOC [18], Pascal Context [58], COCO-Stuff [3], and iSAID [76].”
Transformers in Remote Sensing: A Survey (2022, methodology): “RSP [7] Classification, Segmentation, Detection MillionAID [33], Potsdam [70], iSAID [80], HRSC2016 [81], DOTA [49], CCD [82], LEVIR [61] Overall classification accuracy, mAP, F1 score Investigates pre-training transformers on a large-scale remote sensing dataset.”
RingMo: A Remote Sensing Foundation Model With Masked Image Modeling (2023, methodology): “1) iSAID [37]: The iSAID dataset is a semantic segmentation dataset of DOTA and contains 2806 aerial images with resolutions ranging from 800 × 800 to 4000 × 13 000 pixels.”
Box-supervised Instance Segmentation with Level Set Evolution (2022, methodology): “Our models are trained on train2017 (115K images), and evaluated on val2017 (5K images) and test-dev split (20K images). iSAID [50].”
RemoteCLIP: A Vision Language Foundation Model for Remote Sensing (2023, methodology): “3) Segmentation Data ( SEG-4 ): Four popular remote sensing semantic segmentation datasets, including Vaihin-gen [81], Postdam [82], iSAID [83], and LoveDA [84], are adopted and translated via M2B then B2C.”
D2Det: Towards High Quality Object Detection and Instance Segmentation (2020, methodology): “Our D2Det achieves superior results compared to existing works reported on this dataset [51].”
SkySense: A Multi-Modal Remote Sensing Foundation Model Towards Universal Interpretation for Earth Observation Imagery (2023, methodology): “-S2 dataset [68] and iSAID dataset [78], and change detection with the Dyna.”
Foreground-Aware Relation Network for Geospatial Object Segmentation in High Spatial Resolution Remote Sensing Imagery (2020, background): “iSAID [41] dataset consists of 2,806 HSR remote sensing images.”
Advancing Plain Vision Transformer Toward Remote Sensing Foundation Model (2022, methodology): “Besides the Potsdam2 and iSAID [70] datasets following [10], we also adopt a recently proposed domain adaptation segmentation dataset — LoveDA [71].”