Highlights Metal–organic frameworks (MOFs) are used to directly initiate the gelation of graphene oxide (GO), producing MOF/rGO aerogels. The ultralight magnetic and dielectric aerogels show remarkable microwave absorption performance with ultralow filling contents. Abstract The development of a convenient methodology for synthesizing the hierarchically porous aerogels comprising metal–organic frameworks (MOFs) and graphene oxide (GO) building blocks that exhibit an ultralow density and uniformly distributed MOFs on GO sheets is important for various applications. Herein, we report a facile route for synthesizing MOF/reduced GO (rGO) aerogels based on the gelation of GO, which is directly initiated using MOF crystals. Free metal ions exposed on the surface of MIL-88A nanorods act as linkers that bind GO nanosheets to a three-dimensional porous network via metal–oxygen covalent or electrostatic interactions. The MOF/rGO-derived magnetic and dielectric aerogels Fe3O4@C/rGO and Ni-doped Fe3O4@C/rGO show notable microwave absorption (MA) performance, simultaneously achieving strong absorption and broad bandwidth at low thickness of 2.5 (− 58.1 dB and 6.48 GHz) and 2.8 mm (− 46.2 dB and 7.92 GHz) with ultralow filling contents of 0.7 and 0.6 wt%, respectively. The microwave attenuation ability of the prepared aerogels is further confirmed via a radar cross-sectional simulation, which is attributed to the synergistic effects of their hierarchically porous structures and heterointerface engineering. This work provides an effective pathway for fabricating hierarchically porous MOF/rGO hybrid aerogels and offers magnetic and dielectric aerogels for ultralight MA.

This paper proposes a novel active boundary loss for semantic segmentation. It can progressively encourage the alignment between predicted boundaries and ground-truth boundaries during end-to-end training, which is not explicitly enforced in commonly used cross-entropy loss. Based on the predicted boundaries detected from the segmentation results using current network parameters, we formulate the boundary alignment problem as a differentiable direction vector prediction problem to guide the movement of predicted boundaries in each iteration. Our loss is model-agnostic and can be plugged in to the training of segmentation networks to improve the boundary details. Experimental results show that training with the active boundary loss can effectively improve the boundary F-score and mean Intersection-over-Union on challenging image and video object segmentation datasets.