AI Paper Atlas

Research map / Label-efficient and robust learning / Recognition and long-tail

Cost-Aware Fine-Grained Recognition for IoTs Based on Sequential Fixations

Hanxiao Wang, Venkatesh Saligrama, Stan Sclaroff, Vitaly Ablavsky

ICCV 2019 · Label-efficient and robust learning › Recognition and long-tail · approach: Correspondence and causal representation learning · 2 citations (OpenAlex)

Read the paper

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add a conditioned critic to DDPG (ablation)foveation actor · CUB/Cars/Dogs (foveated)accuracy (%): 51.0 -> 57.1 (CUB); average +8.0%Improved
add coaching from a heuristic oracle policy (ablation)DDPG with conditioned critic · CUB/Cars/Dogs (foveated)accuracy (%): 57.1 -> 67.0 (CUB); average 19.0% gain over DDPGImproved
use full DDPGC3 training (ablation)DRIFT · CUB/Cars/Dogs (foveated)accuracy (%): 67.0 -> 74.4 (CUB); average 25.2% over DDPGImproved
use DRIFT fixations instead of center fixationIoT foveation · CUBaccuracy (%): 59.3 -> 74.4Improved
use saliency or attention fixations instead of center fixationIoT foveation · CUBaccuracy (%): 59.3 -> 37.5 (Saliency), 44.4 (Attention)Hurt
add fallback to full high-acuity image on high entropy (DRIFT_E)DRIFT · Aircraftsaccuracy (%): 86.7 -> 88.0Improved
fuse prediction with DRIFT zoomed attentionInception-V3 · fine-grained standardaccuracy (%): average +1.9%Improved
fuse prediction with DRIFT zoomed attentionResNet-50 · fine-grained standardaccuracy (%): average +2.1%Improved
use DRIFT boxes instead of center boxlocalization · CUB hit ratehit rate (%): 36.2 -> 44.3Improved
use DRIFT boxes instead of center boxlocalization · Aircrafts hit ratehit rate (%): 32.9 -> 50.9Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy