AI Paper Atlas

Research map / Detection and segmentation / CNN features

Object based Scene Representations using Fisher Scores of Local Subspace Projections

Mandar D Dixit, Nuno Vasconcelos

NeurIPS 2016 · Detection and segmentation › CNN features · approach: Local features and context in CNNs · 20 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
use MFA covariance scores (Lambda) instead of GMM variance scores (sigma) (ablation)MFA-FS · MIT Indooraccuracy (%): 53.86 -> 71.11Improved
use MFA covariance scores (Lambda) instead of GMM variance scores (sigma) (ablation)MFA-FS · SUNaccuracy (%): 37.71 -> 53.38Improved
use GMM variance statistics instead of means (ablation)GMM-FV · MIT Indooraccuracy (%): 66.08 -> 53.86Hurt
MFA mean scores instead of GMM mean statistics (ablation)MFA-FS(mu) · MIT Indooraccuracy (%): 66.08 -> 67.68Improved
replace GMM variance FV with MFA subspace-variance FV (ablation)FV · MIT Indooraccuracy (%): 53.86 -> 60.7Improved
combine MFA-FS(mu) with MFA-FS(Lambda) (ablation)MFA-FS · MIT Indooraccuracy (%): ~1%Mixed
concatenate three patch scales instead of best single scale (AlexNet) (ablation)MFA-FS · MIT Indooraccuracy (%): 71.11 -> 73.58Improved
concatenate three patch scales instead of best single scale (VGG-16) (ablation)MFA-FS · SUNaccuracy (%): 61.71 -> 63.31Improved
concatenate three patch scales (VGG-19) instead of 128x128 (ablation)MFA-FS · MIT Indooraccuracy (%): 79.39 -> 81.43Improved
add Places CNN features to MFA-FS (VGG)MFA-FS · MIT Indooraccuracy (%): 81.43 -> 87.23Improved
add Places CNN features to MFA-FS (VGG)MFA-FS · SUNaccuracy (%): 63.31 -> 71.06Improved
task transfer MFA-FS instead of Places CNN dataset transfer (AlexNet)MFA-FS · MIT Indooraccuracy (%): 68.24 -> 73.58Improved

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