Research map / Detection and segmentation / CNN features
SSD: Single Shot MultiBox Detector
Wei Liu, Dragomir Anguelov, Dumitru Erhan, Christian Szegedy, Scott Reed, Cheng-Yang Fu, Alexander C. Berg
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 changed | Where | Measured | Effect |
|---|---|---|---|
| add the extensive data augmentation sampling strategy (ablation) | SSD300 · PASCAL VOC2007 test | mAP: +8.8% mAP | Improved |
| add default boxes with 1/3 and 3 aspect ratios (ablation) | SSD300 · PASCAL VOC2007 test | mAP: -0.6% | Improved |
| add default boxes with 1/2 and 2 aspect ratios (ablation) | SSD300 · PASCAL VOC2007 test | mAP: -2.1% | Improved |
| use atrous subsampled VGG16 instead of full VGG16 (ablation) | SSD300 · PASCAL VOC2007 test | mAP and speed: 74.2 vs 74.3 | No clear effect |
| add prediction from multiple feature layers instead of only conv7 (ablation) | SSD300 · PASCAL VOC2007 test | mAP: 62.4 -> 74.3 | Improved |
| add conv11_2 as a prediction layer (ablation) | SSD300 · PASCAL VOC2007 test | mAP: 74.6 -> 74.3 | No clear effect |
| keep boxes on the image boundary instead of ignoring them (ablation) | SSD300 · PASCAL VOC2007 test | mAP: 63.4 -> 74.3 | Improved |
| add image expansion (zoom out) augmentation (ablation) | SSD300 · PASCAL VOC2007 test (07+12) | mAP: 74.3 -> 77.2 | Improved |
| add image expansion (zoom out) augmentation (ablation) | SSD512 · PASCAL VOC2007 test (07+12) | mAP: 76.8 -> 79.8 | Improved |
| add image expansion (zoom out) augmentation (ablation) | SSD300 · COCO test-dev2015 | mAP@[0.5:0.95]: 23.2 -> 25.1 | Improved |
| increase input size from 300x300 to 512x512 (ablation) | SSD · PASCAL VOC2007 test (07+12) | mAP: 74.3 -> 76.8 | Improved |
| add COCO pretraining before fine-tuning on VOC (ablation) | SSD300 · PASCAL VOC2007 test | mAP: 74.3 -> 79.6 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- Faster R-CNN — Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks (TPAMI 2016) · better on 7 of 7 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Salient Object Detection in the Deep Learning Era: An In-Depth Survey (TPAMI 2021 · similarity 0.86)
- DSOD: Learning Deeply Supervised Object Detectors From Scratch (ICCV 2017 · similarity 0.86)
- Object Detection from Scratch with Deep Supervision (TPAMI 2019 · similarity 0.85)
- CIA-SSD: Confident IoU-Aware Single-Stage Object Detector From Point Cloud (AAAI 2021 · similarity 0.85)
- Generative Sparse Detection Networks for 3D Single-shot Object Detection (ECCV 2020 · similarity 0.85)
- Spatial Self-Distillation for Object Detection with Inaccurate Bounding Boxes (ICCV 2023 · similarity 0.85)
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.
