AI Paper Atlas

Research map / Detection and segmentation / CNN features

Knowledge Diffusion for Distillation

Tao Huang, Yuan Zhang, Mingkai Zheng, Shan You, Fei Wang, Chen Qian, Chang Xu

NeurIPS 2023 · Detection and segmentation › CNN features · approach: Local features and context in CNNs · 23 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 diffusion denoising of student feature before MSE distillationMSE feature KD · ImageNet R18 (R34)Top-1 (%): 70.58 -> 72.22Improved
add diffusion denoising of student feature before MSE distillationMSE feature KD · ImageNet MobileNetV1 (R50)Top-1 (%): 72.39 -> 73.62Improved
replace KL loss with DIST loss in DiffKD logitsDiffKD · ImageNet R18 (R34)Top-1 (%): 72.22 -> 72.49Improved
use DiffKD instead of DISTDIST · ImageNet strong strategies, MobileNetV2Top-1 (%): 74.4 -> 74.9Improved
use DiffKD instead of FGD on COCOFGD · COCO RetinaNet-X101 teacherAP: 40.4 -> 40.7Improved
use DiffKD instead of MasKDMasKD · Cityscapes DeepLabV3-R18mIoU (%): 77.00 -> 77.78Improved
add attention-based MSE loss to DiffKD (ablation)DiffKD · COCO detectionAP: +~0.1 AP (RetinaNet stronger teacher 40.7 -> 41.4)Improved
add adaptive noise matching (ablation)DiffKD · ImageNet MobileNetV1Top-1 (%): 73.34 -> 73.62Improved
distill both feature and logits instead of feature only (ablation)DiffKD · ImageNet MobileNetV1Top-1 (%): 73.16 -> 73.62Improved
add linear autoencoder with 1024 dims (ablation)DiffKD · ImageNet MobileNetV1Top-1 (%): 73.47 -> 73.62Improved
compress autoencoder to 128 dims (ablation)DiffKD · ImageNet MobileNetV1Top-1 (%): 73.47 -> 72.84Hurt
increase denoising steps from 1 to 5 NFEs (ablation)DiffKD · ImageNet MobileNetV1Top-1 (%)Improved

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