Sitemap
Nerd For Tech

NFT is an Educational Media House. Our mission is to bring the invaluable knowledge and experiences of experts from all over the world to the novice. To know more about us, visit https://www.nerdfortech.org/.

Review — GRF-DSOD & GRF-SSD: Improving Object Detection from Scratch via Gated Feature Reuse (Object Detection)

6 min readMar 20, 2021

--

An overview of the proposed GFR-DSOD
@ Medium)

Outline

1. Iterative Feature Re-Utilization

A building block illustrating the iterative feature pyramids

2. Gate-Controlled Adaptive Recalibration

Illustration of the structure of a gate, including: (i) channel-level attention; (ii) global-level attention; and (iii) identity mapping.

2.1. Channel-Level Attention

2.2. Global-Level Attention

2.3. Identity Mapping

3. Feature Reuse for DSOD and SSD

3.1. GRF-DSOD

3.2. GRF-SSD

4. Experimental Results

4.1. Ablation Experiments on PASCAL VOC 2007

Ablation Experiments of gate structure design on PASCAL VOC 2007
Ablation Experiments on PASCAL VOC 2007
Ablation Experiments of SSD300 from scratch on PASCAL VOC 2007

4.2. Results on PASCAL VOC 2007 & 2012

Detection examples on VOC 2007 test set with DSOD / GFR-DSOD models
Convergence Speed Comparison

4.3. Results on MS COCO

Comparisons of two-stage detectors on MS COCO 2015 test-dev set.

--

--

Nerd For Tech
Nerd For Tech

Published in Nerd For Tech

NFT is an Educational Media House. Our mission is to bring the invaluable knowledge and experiences of experts from all over the world to the novice. To know more about us, visit https://www.nerdfortech.org/.

Sik-Ho Tsang
Sik-Ho Tsang

Written by Sik-Ho Tsang

PhD, Researcher. I share what I learn. :) Linktree: https://linktr.ee/shtsang for Twitter, LinkedIn, etc.

No responses yet