Data for "Learning continuous neural representations enables scalable and high-fidelity electron microscopy"

This record contains three electron microscopy datasets used to evaluate ENCODE, a continuous neural representation framework for electron microscopy data compression: STEM_graphene: A simulated HAADF-STEM sequence of graphene vacancy diffusion containing 1,001 frames of 512 × 512 pixels. The dataset was used for compression and temporal-interpolation experiments. TEM_CsPbBr3: An experimental time-resolved TEM sequence of CsPbBr3 nanocrystals containing 500 processed frames of 1,024 × 1,024 pixels, with a temporal sampling interval of 0.12 s. The dataset was used to evaluate the preservation of dynamic structural information. 4DSTEM_Ga2O3: An experimental Ga2O3 4D-STEM datacube with dimensions of 150 × 150 × 256 × 256 pixels. The dataset was used to evaluate the preservation of diffraction features and information required for center-of-mass imaging and ptychographic phase retrieval.