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/

Zendy Plus

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

Zendy Tools

Zendy Open

Recent neural implicit representations (NIRs) have achieved great success inthe tasks of 3D reconstruction and novel view synthesis. However, they requirethe images of a scene from different camera views to be available for one-timetraining. This is expensive especially for scenarios with large-scale scenesand limited data storage. In view of this, we explore the task of incrementallearning for NIRs in this work. We design a student-teacher framework tomitigate the catastrophic forgetting problem. Specifically, we iterate theprocess of using the student as the teacher at the end of each time step andlet the teacher guide the training of the student in the next step. As aresult, the student network is able to learn new information from the streamingdata and retain old knowledge from the teacher network simultaneously. Althoughintuitive, naively applying the student-teacher pipeline does not work well inour task. Not all information from the teacher network is helpful since it isonly trained with the old data. To alleviate this problem, we further introducea random inquirer and an uncertainty-based filter to filter useful information.Our proposed method is general and thus can be adapted to different implicitrepresentations such as neural radiance field (NeRF) and neural surface field.Extensive experimental results for both 3D reconstruction and novel viewsynthesis demonstrate the effectiveness of our approach compared to differentbaselines.

UNIKD: UNcertainty-filtered Incremental Knowledge Distillation for Neural Implicit Representation