Table 3 Table showing deep learning based fusion techniques and the training datasets used.
Technique | Training datasets | |
|---|---|---|
DenseFuse | Microsoft Common Objects in Context (MS-COCO)10 | 79,000 inputs 1000 validation images |
FusionGAN | TNO5 | 64,381 pairs |
IFCNN | Online generated from ImageNet dataset, NYU-D2 dataset12 | Nearly 100,000 image pairs |
PIAFusion | Multi-Spectral Road Scenarios (MSRS)9, MFNet dataset8 | Full MSRS dataset |
PMGI | TNO5, HarvardMed2 | 124 image pairs, 407,696 patch pairs |
RFN-Nest | Microsoft Common Objects in Context (MS-COCO)10, Korea Advanced Institute of Science and Technology (KAIST)11 | Nearly 80,000 images |
SDNet | TNO5, HarvardMed2 | Nearly, 370,839 patch pairs |
SeAFusion | MFNet8 | Whole MFNet dataset with patches |
U2Fusion | RoadScene1, Harvard2, Lytro4, TNO5, EMPAHDR6, FLIR7 | All imaged in the datasets |
DLIEP | MS-COCO10 | 82,612 images (MS-COCO) |
PPTFusion | MS-COCO10 | 82,612 images (MS-COCO) |
