IEEE Geoscience and Remote Sensing Magazine - December 2021 - 275

high-resolution (1-m GSD) land
cover maps to identify semantic
changes between 2013 and 2017.
The performance was evaluated
using the intersection-overunion
metric averaged over eight types
of change. The challenge was twofold:
to detect which parts of the
image changed between two highresolution
aerial images, and to
identify the class of change based
on weak supervision.
The 2021 DFC tackled two fundamental
technical challenges rooted
in real social problems: 1) analysis
of multisensor, multiresolution, and
multitemporal data, and 2) learning
from weak supervision. These two
issues are major challenges in a wide
range of fields, from Earth observation
to computer vision and machine
learning. The most important
feature of the 2021 DFC is that it is
directly related to social issues such
as energy equality and environmental conservation. The results
of the contest will have a significant impact, not only
in terms of technological development, but also as a tool
for solving real social problems.
OUTCOME OF THE CONTEST
The first- to fourth-ranked teams in each track were awarded
as winners of the contest and presented their solutions
during the 2021 IEEE International Geoscience and Remote
Sensing Symposium (IGARSS 2021).
The four winning teams of the 2021 DFC Track DSE were
the following:
◗ First place: fengkexin team; Yanbiao Ma, Yuxin Li, Kexin
Feng, and Xueli Geng (Xidian University, China); dualtask
models based on squeeze-and-excitation networks
followed by postprocessing based on expert priors [8].
◗ Second place: WHU_YuXia team; Yu Xia, Qi Huang, and
Hongyan Zhang (Wuhan University, China); ensembling
of single-task and dual-task models based on global
context convolutional neural networks and random
forests [9].
FIGURE 1. The awards for DFC 2021 were handed out during the virtual award ceremony of
IGARSS21. The picture shows representatives of all eight winning teams as well as the chair
of IADF.
◗ Third place: dimartinot team; Thomas Di Martino, Maxime
Lenormand, and Elise Colin Koeniguer (ONERA,
France); multibranch CNN with 3D convolutions and
EfficientNet [10].
◗ Third place: JIOJIO team; Ruoxian Feng, Mengjiao
Wang, Xuanming Zhang, and Jun Zhang (Xidian University,
China); ensembling of UNet and LinkNet with
depthwise overparameterized convolutional layers [11].
The four winning teams of the 2021 DFC Track MSD
were the following:
◗ First place: AsheLee team; Zhuohong Li, Fangxiao Lu,
Hongyan Zhang, Guangyi Yang, and Liangpei Zhang
(Wuhan University, China) [12].
◗ Second place: tulilin team; Lilin Tu, Jiayi Li, and Xin
Huang (Wuhan University, China) [13].
◗ Third place: baoqianyue team; Qianyue Bao, Yang Liu,
Zixiao Zhang, Dafan Chen, Yuting Yang, Licheng Jiao,
and Fang Liu (Xidian University, China) [14].
◗ Fourth place: EVER team; Zhuo Zheng, Yinhe Liu, Shiqi
Tian, Junjue Wang, Ailong Ma, and Yanfei Zhong (Wuhan
University, China) [15].
DECEMBER 2021 IEEE GEOSCIENCE AND REMOTE SENSING MAGAZINE
275
IEEE Geoscience and Remote Sensing Magazine - December 2021

Table of Contents for the Digital Edition of IEEE Geoscience and Remote Sensing Magazine - December 2021
