AI Paper Atlas

Research map / Detection and segmentation / CNN features

Delving Into Robust Object Detection From Unmanned Aerial Vehicles: A Deep Nuisance Disentanglement Approach

Zhenyu Wu, Karthik Suresh, Priya Narayanan, Hongyu Xu, Heesung Kwon, Zhangyang Wang

ICCV 2019 · Detection and segmentation › CNN features · approach: Box regression and loss design · 99 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

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 altitude nuisance disentanglement (gamma1=0.03) (ablation)Faster-RCNN · UAVDT (vehicle)AP: 45.64 -> 45.92Improved
add view nuisance disentanglement (gamma2=0.01) (ablation)Faster-RCNN · UAVDT (vehicle)AP: 45.64 -> 46.16Improved
add weather nuisance disentanglement (gamma3=0.01) (ablation)Faster-RCNN · UAVDT (vehicle)AP: 45.64 -> 46.62Improved
raise altitude gamma1 from 0.03 to 0.05 (ablation)NDFT altitude only · UAVDT (vehicle)AP: 45.92 -> 44.36Hurt
raise weather gamma3 from 0.01 to 0.1 (ablation)NDFT weather only · UAVDT (vehicle)AP: 46.62 -> 43.60Hurt
add weather disentanglement on night images (ablation)Faster-RCNN · UAVDT (night)AP: 52.14 -> 59.66Improved
combine altitude and view disentanglement (ablation)NDFT · UAVDT (vehicle)AP: 46.16 -> 46.88Improved
add weather to altitude and view disentanglement (ablation)NDFT A+V · UAVDT (vehicle)AP: 46.88 -> 47.91Improved
replace Faster-RCNN with FPN backbone in NDFTNDFT A+V+W · UAVDT (vehicle)AP: 49.05 -> 52.03Improved
use adversarial nuisance loss instead of auxiliary multi-task loss (ablation)NDFT · UAVDT (vehicle)AP: 45.69 -> 46.81Improved
add NDFT to detections used by SORT trackerSORT tracking · UAVDT MOTIDF: 43.7 -> 52.9Improved
add NDFT to detections used by SORT trackerSORT tracking · UAVDT MOTMOTA: 39.0 -> 38.4Hurt

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