AI Paper Atlas

Research map / Video / Video segmentation and tracking

Joint Detection and Recounting of Abnormal Events by Learning Deep Generic Knowledge

Ryota Hinami, Tao Mei, Shin'ichi Satoh

ICCV 2017 · Video › Video segmentation and tracking · approach: Weakly supervised video anomaly detection · 278 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
replace HOG features with object-trained Fast R-CNN features (ablation)NN anomaly detector · Unseen object detection (COCO)mAP (%): 8.3 -> 20.6Improved
replace HOG features with action-trained Fast R-CNN features (ablation)NN anomaly detector · Unseen action detection (COCO)mAP (%): 16.8 -> 30.2Improved
replace HOG features with Fast R-CNN features (ablation)NN anomaly detector · Ped2 and Avenue17 ROCROC curveImproved
use multi-task instead of action-only Fast R-CNN (ablation)anomaly detector · Avenue17 (frame-level)AUC (%): 89.8 -> 89.2Hurt
use multi-task instead of action-only Fast R-CNN (ablation)anomaly detector · UCSD Ped2 (frame-level)AUC (%): 92.2 -> 90.8Hurt
use multi-task instead of object-only Fast R-CNN (ablation)anomaly detector · Unseen object detection (COCO)mAP (%): 20.6 -> 16.9Hurt
use multi-task instead of object-only Fast R-CNN for action concepts (ablation)anomaly detector · Unseen action detection (COCO)mAP (%): 20.5 -> 29.6Improved
swap anomaly detector among NN, OC-SVM and KDE (ablation)FRCN features · Avenue17 and Ped2AUCNo clear effect
train SDAE on normal data instead of generic data (ablation)SDAE · Unseen object detection (COCO)mAP (%): 8.1 -> 10.9Improved
use MT-FRCN features instead of HOG with SVM (ablation)recounting · Unseen object recounting (COCO)AUC: 2.1 -> 24.7Improved
use KDE on MT-FRCN output instead of SVM scores (ablation)recounting · Unseen object recounting (COCO)AUC: 24.7 -> 26.8Improved
use MT-FRCN output instead of SVM scores (ablation)recounting · Unseen action recounting (COCO)AUC: 16.2 -> 16.5No clear effect

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