Expandable-RCNN: toward high-efficiency incremental few-shot object detection.

This study aims at addressing the challenging incremental few-shot object detection (iFSOD) problem toward online adaptive detection. iFSOD targets to learn novel categories in a sequential manner, and eventually, the detection is performed on all learned categories. Moreover, only a few training samples are available for all sequential novel classes in these situations. In this study, we propose an efficient yet suitably simple framework, Expandable-RCNN, as a solution for the iFSOD problem, which allows online sequentially adding new classes with zero retraining of the base network. We achieve this by adapting the Faster R-CNN to the few-shot learning scenario with two elegant components to effectively address the overfitting and category bias. First, an IOU-aware weight imprinting strategy is proposed to directly determine the classifier weights for incremental novel classes and the background class, which is with zero training to avoid the notorious overfitting issue in few-shot learning. Second, since the above zero-retraining imprinting approach may lead to undesired category bias in the classifier, we develop a bias correction module for iFSOD, named the group soft-max layer (GSL), that efficiently calibrates the biased prediction of the imprinted classifier to organically improve classification performance for the few-shot classes, preventing catastrophic forgetting. Extensive experiments on MS-COCO show that our method can significantly outperform the state-of-the-art method ONCE by 5.9 points in commonly encountered few-shot classes.

High Density Glycopolymers Functionalized Perylene Diimide Nanoparticles for Tumor-Targeted Photoacoustic Imaging and Enhanced Photothermal Therapy.

Near-infrared (NIR) absorbing nanoagents with functions of photoacoustic imaging (PAI) and photothermal therapy (PTT) have received great attention for cancer therapy. However, endowing them with multifunctions, especially targeting ability, for enhancing in vivo PAI/PTT generally suffers from the problems of synthetic complexity and low surface density of function groups. We herein report high density glycopolymers coated perylenediimide nanoparticles (PLAC-PDI NPs), self-assembled by poly(lactose)-modified perylenediimide (PLAC-PDI), as tumor-targeted PAI/PTT nanoagents. Atom transfer radical polymerization and click reaction were used in sequence to prepare PLAC-PDI, which can accurately control the content of poly(lactose) (PLAC) in PLAC-PDI and endow PLAC-PDI NPs with high density PLAC surface. The high density PLAC surface provided NPs with long-time colloidal stability, outstanding stability in serum and light, and specific targeting ability to cancer cells and tumors. Meanwhile, PLAC-PDI NPs also presented high photothermal conversion efficiency of 42% by virtue of strong π-π interactions among perylenediimide molecules. In living mice, PAI experiments revealed that PLAC-PDI NPs exhibited effective targeting ability and enhanced PTT efficacy to HepG2 tumor compared with control groups, lactose blocking, and ASGP-R negative tumor groups. Overall, our work provids new insights for designing glycopolymers-based therapeutic nanoagents for efficient tumor imaging and antitumor therapy.