Remote Sensing-based Multi-view Learning (MVL) Models that perform Data Fusion

May 5, 2026 ยท View on GitHub

List created based on our work Common Practices and Taxonomy in Deep Multiview Fusion for Remote Sensing Applications

Foundational models can be found at TorchGeo or HuggingFace

List of MVL models proposed to learn from remote sensing (RS) multi-view data. :satellite: :earth_americas: :satellite:

NameReferenceDescriptionCode
MVLMena et al. 2025Multiple fusion strategies with temporal datahttps://github.com/fmenat/optimal-multiview-crop-classifier
SSL4EO-S12Wang et al. 2024Self-supervised modelhttps://github.com/zhu-xlab/SSL4EO-S12
EarthGPTZhang et al. 2024Feature-level fusion with transformer layershttps://github.com/wivizhang/EarthGPT
ContextFormerBenson et al. 2024Feature-level fusion with transformerhttps://github.com/vitusbenson/greenearthnet
SEnSeIv2Francis 2024Sensor-invariant modelhttps://github.com/aliFrancis/SEnSeIv2
OmniSatAstruc et al. 2024Feature-level fusion with transformer layers and pre-traininghttps://github.com/gastruc/OmniSat
MMEarthNedungadi et al. 2024Single-view input to predict another views (pretext task)https://github.com/vishalned/MMEarth-train
MambaDifussionDu et al. 2024Feature level fusion with Mamba over different layers (skip connection) including diffusion modelshttps://github.com/WenliangDu/MambaDiffusion
FusionMambaPeng et al. 2024Dense fusion with Mamba layershttps://github.com/PSRben/FusionMamba
SatViTFuller et al.Input-level fusion with ViT (with self-supervised training)https://github.com/antofuller/SatViT
ELECTSRusswurm et al. 2023Input-level fusion with LSTM.https://github.com/marccoru/elects
MV CNNFerrari et al. 2023Multiple fusion strategies with 2D CNN (encoder-decoder)https://github.com/felferrari/deforestation-from-data-fusion
AFCF3D-NetYe et al. 2023Input-level fusion with 3D CNN.https://github.com/wm-Githuber/AFCF3D-Net
UnCRtainTSEbel et al 2023Input fusion with 2D CNN and attention.https://github.com/PatrickTUM/UnCRtainTS
MFTRoy et al. 2023Feature-level fusion with transformer modules (one source - LIDAR- is used as a query over the main source - optical)https://github.com/AnkurDeria/MFT
PRESTOTseng et al. 2023Input-level fusion with transformer modules (self-supervised pretraining)https://github.com/nasaharvest/presto
Cross-HLRoy et al. 2023Feature-level fusion with directed attention in transformer layershttps://github.com/AtriSukul1508/Cross-HL
SCT FusionHoffman et al. 2023Dense fusion with tranformer layers and class tokens in allhttps://git.tu-berlin.de/rsim/sct-fusion
MMST-ViTLin et al. 2023Feature-level fusion with transformer layershttps://github.com/fudong03/MMST-ViT
DiffusionSatKhanna et al. 2023Multi-modal diffusion generative modelhttps://github.com/samar-khanna/DiffusionSat
AMM-FuseNetMa et al. 2022Feature-level fusion with 2D CNN and attention.https://github.com/oktaykarakus/ReSIF/tree/main/AMM-FuseNet
MCANetLi et al. 2022Dense fusion with 2D CNN and cross attention.https://github.com/yisun98/SOLC
ChangeFormerBandara et al. 2022Dense fusion with transformer and attention.https://github.com/wgcban/ChangeFormer
CMAFFQingyun et al. 2022Dense fusion with 2D CNN and cross attention.https://github.com/DocF/CMAFF
OmbriaNetDrakonakis et al. 2022Feature fusion with 2D CNN and skip-connectionshttps://github.com/geodrak/OMBRIA
DCSA-NetWang et al. 2022Hybrid fusion with 2D CNN and attention.https://github.com/Julia90/DCSA-Net
Siamese U-NetCummings et al. 2022Dense fusion with 2D CNN and skip-connectionshttps://github.com/solcummings/earthvision2021-weakly-supervised
AM3^3NetWang et al. 2022Feature-level fusion with 2D CNN and cross attention.https://github.com/Cimy-wang/AM3Net_Multimodal_Data_Fusion
MAHiDFNetWang et al. 2022Dense feature fusion with 2D CNN.https://github.com/SYFYN0317/-MAHiDFNet
EndNetHong et al. 2022Feature-level fusion with 2D CNN and view-reconstruction.https://github.com/danfenghong/IEEE_GRSL_EndNet
SE2^2NetFang et al. 2022Feature-level fusion with 2D CNN.https://github.com/likyoo/Multimodal-Remote-Sensing-Toolkit
MV CNNLu et al. 2022Feature-level fusion with 2D CNN and adaptive attention.https://github.com/GeoX-Lab/UnifiedDL-UFZ-extraction
IP-CNNZhang et al. 2022Feature-level fusion with 2D CNN and view-reconstruction.https://github.com/HelloPiPi/IP-CNN-code
ASF2NGao et al. 2022Feature-level fusion with 2D CNN and attention.https://github.com/zhonghaocheng/ELSEVIER_IJAEOG_AS2F2N << Empty code
SEnSeIFrancis et al. 2022Sensor invariant model based on 2D CNNhttps://github.com/aliFrancis/SEnSeI
MV NNDanilevicz et al. 2021Feature-level fusion with tabular NN and 2D CNN.https://github.com/mdanilevicz/maize_early_yield_prediction
CFCNNHe et al. 2021Feature-level fusion with 2D and 1D CNN.https://github.com/SysuHe/MultiSourceData_CFCNN
S2FLHong et al. 2021Feature-level fusion with feature constraints.https://github.com/danfenghong/ISPRS_S2FL
CMGFNetHosseinpour et al. 2022Dense fusion with 2D CNN and gated attention.https://github.com/hamidreza2015/CMGFNet-Building_Extraction
MDL-RSHong et al. 2021Multiple fusion strategies with NN.https://github.com/danfenghong/IEEE_TGRS_MDL-RS
MV PSE-TAEOfori-Ampofo et al. 2021Multiple fusion strategies with PSE-TAE.https://github.com/ellaampy/CropTypeMapping
CCR-NetWu et al. 2021Feature-level fusion with 2D CNN and cross-view reconstruction.https://github.com/danfenghong/IEEE_TGRS_CCR-Net
LFMC from SARRao et alInput-level fusion with LSTMhttps://github.com/kkraoj/lfmc_from_sar
FusAtNetMohla et al. 2020Feature-level fusion with 2D CNN and cross attention.https://github.com/ShivamP1993/FusAtNet
HRWNZhao et al. 2020Input-level fusion with 2D CNN and pixel graph constraints.https://github.com/xudongzhao461/HRWN
UNet-CLSTMRustowicz et al. 2019Decision-level fusion with 2D CNN and convolutional-LSTMhttps://github.com/roserustowicz/crop-type-mapping
Multi3NetRudner et al. 2019Feature-level fusion with 2D CNN.https://github.com/FrontierDevelopmentLab/multi3net
V-FuseNetAudebert et al. 2018Dense fusion with 2D CNN and central model.https://github.com/nshaud/DeepNetsForEO
MV CNNXu et al. 2018Feature-level fusion with 2D CNN.https://github.com/Hsuxu/Two-branch-CNN-Multisource-RS-classification << Not available

Feel free to ask me to update some content!

List of MVL models robust to missing data and views

NameReferenceDescriptionCode
MDiCoMena et al. 2025Decision-level co-learning to handle single-view predictionhttps://github.com/fmenat/MDiCo
DSensD+Mena et al. 2025Decision-level fusion with sensor dropout and mutual distillationhttps://github.com/fmenat/dsensdp
FCoM-*Mena et al. 2025Feature-level fusion with all combinations of missing views and dynamic fusionhttps://github.com/fmenat/CoM-views
OOD FusionGawlikowski et al. 2023Input-level, Feature-level, and Decision-level fusion with CNN and weighted average aggregationhttps://github.com/JakobCode/OOD_DataFusion

Feel free to ask me to update some content!

Some abbreviations

Abbrevationname
CNNconvolutional neural network
LSTMlong-short term memory
NNneural network
PSE-TAEpixel set encoder - temporal attention encoder