English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Privacy and memory are two recurring themes in a broad conversation about thesocietal impact of AI. These concerns arise from the need for huge amounts ofdata to train deep neural networks. A promise of Generalized Few-shot ObjectDetection (G-FSOD), a learning paradigm in AI, is to alleviate the need forcollecting abundant training samples of novel classes we wish to detect byleveraging prior knowledge from old classes (i.e., base classes). G-FSODstrives to learn these novel classes while alleviating catastrophic forgettingof the base classes. However, existing approaches assume that the base imagesare accessible, an assumption that does not hold when sharing and storing datais problematic. In this work, we propose the first data-free knowledgedistillation (DFKD) approach for G-FSOD that leverages the statistics of theregion of interest (RoI) features from the base model to forge instance-levelfeatures without accessing the base images. Our contribution is three-fold: (1)we design a standalone lightweight generator with (2) class-wise heads (3) togenerate and replay diverse instance-level base features to the RoI head whilefinetuning on the novel data. This stands in contrast to standard DFKDapproaches in image classification, which invert the entire network to generatebase images. Moreover, we make careful design choices in the novel finetuningpipeline to regularize the model. We show that our approach can dramaticallyreduce the base memory requirements, all while setting a new standard forG-FSOD on the challenging MS-COCO and PASCAL-VOC benchmarks.

NIFF: Alleviating Forgetting in Generalized Few-Shot Object Detection via Neural Instance Feature Forging