特別感謝實驗室小雷同學匯總此篇，日后學習目標跟蹤可以有個好的方向好的借鑒，哪怕是比賽的時候選模型都可以參考一下。
----------------------------------------------------------
| 論文對應序號 |
method |
dataset |
code |
||
|
|
|
VOC2007 |
VOC2012 |
COCO |
|
| 1 |
Cascade R-CNN |
|
|
42.8(AP) |
有 |
| 2 |
Relation Net |
|
|
39.0(加到別的方法上) |
有 |
| 3 |
RefineDet |
85.8 |
86.8 |
41.8（AP） |
有 |
| 4 |
SNIP |
|
|
|
有 |
| 5 |
R-FCN-3000 |
43.3(ImageNet) |
無 |
||
| 6 |
DES |
84.3 |
83.7 |
32.8 |
無 |
| 7 |
STDN |
80.9 |
|
31.8 |
有 |
| 8 |
W2F |
52.4 |
47.8 |
|
無 |
| 9 |
無簡寫 |
51.2 |
|
|
無 |
| 10 |
MELM |
47.3 |
42.4 |
|
有 |
| 11 |
SSM |
62.9 |
|
|
有 |
| 12 |
無簡寫 |
82.9 |
|
35.6(AP) |
有 |
| 13 |
PAD |
80.7 |
79.5 |
|
無 |
| 14 |
ZLDN |
47.6 |
42.9 |
|
無 |
| 15 |
無簡寫 |
|
|
39.5 |
無 |
| 16 |
MegDet |
|
|
52.5(mmAP) |
無 |
| 17 |
drl-RPN |
76.4 |
72.2 |
|
有 |
| 18 |
SIN |
76.0 |
73.1 |
23.2(AP) |
有 |
| 19 |
SOD-MTGAN |
|
|
41.4(AP) |
無 |
| 20 |
ML-LocNet |
49.7 |
43.6 |
16.2(COCO2014) |
無 |
| 21 |
DetNet |
|
|
40.3 |
有 |
| 22 |
無簡寫 |
50.4 |
69.3 |
|
無 |
| 23 |
無簡寫 |
25.4 |
22.9 |
|
無 |
| 24 |
無簡寫 |
82.4 |
81.1 |
34.6(AP) |
無 |
| 25 |
RFB-NET |
82.2 |
|
29.7(COCO2014) 34.4(COCO2015) |
有 |
| 26 |
PFP-NET |
84.1 |
83.7 |
41.8 |
有 |
| 27 |
TS2C |
44.3 |
40.0 |
|
無 |
| 28 |
SAN |
82.8 |
|
43.4 |
無 |
| 29 |
無簡寫 |
|
81.2 |
mmAP:39.3(COCO2017) |
無 |
| 30 |
無簡寫 |
|
|
42.0(AP) |
無 |
附：
(1)論文對應序號中，序號1-18篇收錄於CVPR，19-30收錄於ECCV。
(2)在經典數據庫的檢測精度取在論文中實現的最高精度，不考慮base network。
(3)method列僅寫出算法簡稱。
(4)針對COCO數據集的檢測結果不可進行統一比較。有的是在COCO2014、COCO2015或者是COCO2017上測試，評價指標稍有不同。
(5)CVPR2019論文未公布。
======以下排名僅對論文中有在對應數據集測試的算法進行排序=========
VOC2007數據集排名
| 論文對應序號 |
method |
mAP |
排名 |
| 3 |
RefineDet |
85.8 |
1 |
| 6 |
DES |
84.3 |
2 |
| 26 |
PFP-NET |
84.1 |
3 |
| 12 |
無簡寫 |
82.9 |
4 |
| 28 |
SAN |
82.8 |
5 |
| 24 |
無簡寫 |
82.4 |
6 |
| 25 |
RFB-NET |
82.2 |
7 |
| 7 |
STDN |
80.9 |
8 |
| 13 |
PAD |
80.7 |
9 |
| 17 |
drl-RPN |
76.4 |
10 |
| 18 |
SIN |
76.0 |
11 |
| 11 |
SSM |
62.9 |
12 |
| 8 |
W2F |
52.4 |
13 |
| 9 |
無簡寫 |
51.2 |
14 |
| 22 |
無簡寫 |
50.4 |
15 |
| 20 |
ML-LocNet |
49.7 |
16 |
| 14 |
ZLDN |
47.6 |
17 |
| 10 |
MELM |
47.3 |
18 |
| 27 |
TS2C |
44.3 |
19 |
| 23 |
無簡寫 |
25.4 |
20 |
VOC2012數據集排名
| 論文對應序號 |
method |
mAP |
排名 |
| 3 |
RefineDet |
86.8 |
1 |
| 6 |
DES |
83.7 |
2 |
| 26 |
PFP-NET |
83.7 |
2 |
| 29 |
無簡寫 |
81.2 |
3 |
| 24 |
無簡寫 |
81.1 |
4 |
| 13 |
PAD |
79.5 |
5 |
| 18 |
SIN |
73.1 |
6 |
| 17 |
drl-RPN |
72.2 |
7 |
| 22 |
無簡寫 |
69.3 |
8 |
| 8 |
W2F |
47.8 |
9 |
| 20 |
ML-LocNet |
43.6 |
10 |
| 14 |
ZLDN |
42.9 |
11 |
| 10 |
MELM |
42.4 |
12 |
| 27 |
TS2C |
40.0 |
13 |
| 23 |
無簡寫 |
22.9 |
14 |
| 22 |
無簡寫 |
50.4 |
15 |
| 20 |
ML-LocNet |
49.7 |
16 |
| 14 |
ZLDN |
47.6 |
17 |
| 10 |
MELM |
47.3 |
18 |
| 27 |
TS2C |
44.3 |
19 |
| 23 |
無簡寫 |
25.4 |
20 |
COCO數據集排名
| 論文對應序號 |
method |
mAP |
排名 |
| 16 |
MegDet |
52.5(mmAP) |
1 |
| 28 |
SAN |
43.4 |
2 |
| 1 |
Cascade R-CNN |
42.8(AP) |
3 |
| 30 |
無簡寫 |
42.0(AP) |
4 |
| 26 |
PFP-NET |
41.8 |
5 |
| 3 |
RefineDet |
41.8（AP） |
6 |
| 19 |
SOD-MTGAN |
41.4(AP) |
7 |
| 21 |
DetNet |
40.3 |
8 |
| 15 |
無簡寫 |
39.5 |
9 |
| 29 |
無簡寫 |
mmAP:39.3(COCO2017) |
10 |
| 2 |
Relation Net |
39.0(加到別的方法上) |
11 |
| 12 |
無簡寫 |
35.6(AP) |
12 |
| 24 |
無簡寫 |
34.6(AP) |
13 |
| 25 |
RFB-NET |
29.7(COCO2014) 34.4(COCO2015) |
14 |
| 6 |
DES |
32.8 |
15 |
| 7 |
STDN |
31.8 |
16 |
| 18 |
SIN |
23.2(AP) |
17 |
| 20 |
ML-LocNet |
16.2(COCO2014) |
18 |
1、Cascaded RCNN
| 論文 |
Cascade R-CNN : Delving into High Quality Object Detection |
| 論文鏈接 |
https://arxiv.org/abs/1712.00726 |
| 代碼鏈接 |
https://github.com/zhaoweicai/cascade-rcnn |
實驗結果
2、Relation Net
| 論文 |
Relation Networks for Object Detection |
| 論文鏈接 |
https://arxiv.org/abs/1711.11575 |
| 代碼鏈接 |
https://github.com/msracver/Relation-Networks-for-Object-Detection |
實驗結果
（實驗是針對two stage系列的目標檢測算法而言，在ROI Pooling后的兩個全連接層和NMS模塊引入object relation module，如Figure1所示，因此做到了完整的end-to-end訓練。）
3、RefineDet
| 論文 |
Single-Shot Refinement Neural Network for Object Detection |
| 論文鏈接 |
https://arxiv.org/abs/1711.06897 |
| 代碼鏈接 |
https://github.com/sfzhang15/RefineDet |
實驗結果
4、SNIP
| 論文 |
An Analysis of Scale Invariance in Object Detection – SNIP |
| 論文鏈接 |
https://arxiv.org/abs/1711.08189 |
| 代碼鏈接 |
http://bit.ly/2yXVg4c（打不開） |
實驗結果
5、R-FCN-3000
| 論文 |
R-FCN-3000 at 30fps: Decoupling Detection and Classification |
| 論文鏈接 |
https://arxiv.org/abs/1712.01802 |
| 代碼鏈接 |
|
ImageNet實驗結果
6、DES
| 論文 |
Single-Shot Object Detection with Enriched Semantics |
| 論文鏈接 |
https://arxiv.org/abs/1712.00433 |
| 代碼鏈接 |
|
實驗結果
7、STDN
| 論文 |
Scale-Transferrable Object Detection |
| 論文鏈接 |
http://openaccess.thecvf.com/content_cvpr_2018/papers/Zhou_Scale-Transferrable_Object_Detection_CVPR_2018_paper.pdf |
| 代碼鏈接 |
https://github.com/arvention/STDN |
實驗結果
8、W2F
| 論文 |
W2F: A Weakly-Supervised to Fully-Supervised Framework for Object Detection |
| 論文鏈接 |
http://openaccess.thecvf.com/content_cvpr_2018/papers/Zhang_W2F_A_Weakly-Supervised_CVPR_2018_paper.pd |
| 代碼鏈接 |
|
實驗結果
9、
| 論文 |
Multi-Evidence Filtering and Fusion for Multi-Label Classification, Object Detection and Semantic Segmentation Based on Weakly Supervised Learning |
| 論文鏈接 |
http://openaccess.thecvf.com/content_cvpr_2018/papers/Ge_Multi-Evidence_Filtering_and_CVPR_2018_paper.pdf |
| 代碼鏈接 |
|
實驗結果
10、MELM
| 論文 |
Min-Entropy Latent Model for Weakly Supervised Object Detection |
| 論文鏈接 |
http://openaccess.thecvf.com/content_cvpr_2018/papers/Wan_Min-Entropy_Latent_Model_CVPR_2018_paper.pdf |
| 代碼鏈接 |
https://github.com/Winfrand/MELM |
實驗結果
11、SSM
| 論文 |
Towards Human-Machine Cooperation: Self-supervised Sample Mining for Object Detection |
| 論文鏈接 |
http://openaccess.thecvf.com/content_cvpr_2018/papers/Wang_Towards_Human-Machine_Cooperation_CVPR_2018_paper.pdf |
| 代碼鏈接 |
https://github.com/yanxp/SSM-Pytorch |
實驗結果
12、
| 論文 |
Feature Selective Networks for Object Detection |
| 論文鏈接 |
https://arxiv.org/abs/1711.08879 |
| 代碼鏈接 |
https://github.com/robwec/feature-selective-networks |
實驗結果
13、PAD
| 論文 |
Pseudo Mask Augmented Object Detection |
| 論文鏈接 |
http://openaccess.thecvf.com/content_cvpr_2018/papers/Zhao_Pseudo_Mask_Augmented_CVPR_2018_paper.pdf |
| 代碼鏈接 |
|
實驗結果
14、ZLDN
| 論文 |
Zigzag Learning for Weakly Supervised Object Detection |
| 論文鏈接 |
http://openaccess.thecvf.com/content_cvpr_2018/papers/Zhang_Zigzag_Learning_for_CVPR_2018_paper.pdf |
| 代碼鏈接 |
|
實驗結果
15、
| 論文 |
Learning Globally Optimized Object Detector via Policy Gradient |
| 論文鏈接 |
http://openaccess.thecvf.com/content_cvpr_2018/papers/Rao_Learning_Globally_Optimized_CVPR_2018_paper.pdf |
| 代碼鏈接 |
|
實驗結果
16、MegDet
| 論文 |
MegDet: A Large Mini-Batch Object Detector |
| 論文鏈接 |
http://openaccess.thecvf.com/content_cvpr_2018/papers/Peng_MegDet_A_Large_CVPR_2018_paper.pdf |
| 代碼鏈接 |
|
實驗結果
The MegDet is the backbone of our submission (mmAP 52.5%) to COCO 2017 Challenge, where we won the 1st place of Detection task.
17、drl-RPN
| 論文 |
Deep Reinforcement Learning of Region Proposal Networks for Object Detection |
| 論文鏈接 |
http://openaccess.thecvf.com/content_cvpr_2018/papers/Pirinen_Deep_Reinforcement_Learning_CVPR_2018_paper.pdf |
| 代碼鏈接 |
https://github.com/aleksispi/drl-rpn-tf |
實驗結果
18、SIN
| 論文 |
Structure Inference Net: Object Detection Using Scene-Level Context and Instance-Level Relationships |
| 論文鏈接 |
http://openaccess.thecvf.com/content_cvpr_2018/papers/Liu_Structure_Inference_Net_CVPR_2018_paper.pdf |
| 代碼鏈接 |
https://github.com/choasup/SIN |
實驗結果
以下是ECCV2018論文
19、SOD-MTGAN
論文：SOD-MTGAN: Small Object Detection via Multi-Task Generative Adversarial Network
論文鏈接：
http://openaccess.thecvf.com/content_ECCV_2018/papers/Yongqiang_Zhang_SOD-MTGAN_Small_Object_ECCV_2018_paper.pdf
代碼鏈接：
實驗結果
20、ML-LocNet
論文：ML-LocNet: Improving Object Localization with Multi-view Learning Network
論文鏈接：
http://openaccess.thecvf.com/content_ECCV_2018/papers/Xiaopeng_Zhang_ML-LocNet_Improving_Object_ECCV_2018_paper.pdf
代碼鏈接：
實驗結果
21、DetNet
論文：DetNet: Design Backbone for Object Detection
論文鏈接：
http://openaccess.thecvf.com/content_ECCV_2018/papers/Zeming_Li_DetNet_Design_Backbone_ECCV_2018_paper.pdf
代碼鏈接：https://github.com/guoruoqian/DetNet_pytorch
或者https://github.com/BigDeviltjj/mxnet-detnet
實驗結果
22、
論文：Weakly Supervised Region Proposal Network and Object Detection
論文鏈接：
http://openaccess.thecvf.com/content_ECCV_2018/papers/Peng_Tang_Weakly_Supervised_Region_ECCV_2018_paper.pdf
代碼鏈接：
實驗結果
23、
論文：Zero-Annotation Object Detection with Web Knowledge Transfer
論文鏈接：
http://openaccess.thecvf.com/content_ECCV_2018/papers/Qingyi_Tao_Zero-Annotation_Object_Detection_ECCV_2018_paper.pdf
代碼鏈接：
實驗結果
24、
論文：Deep Feature Pyramid Reconfiguration for Object Detection
論文鏈接：
http://openaccess.thecvf.com/content_ECCV_2018/papers/Tao_Kong_Deep_Feature_Pyramid_ECCV_2018_paper.pdf
代碼鏈接：
實驗結果
25、RFB-NET
論文：Receptive Field Block Net for Accurate and Fast Object Detection
論文鏈接：
http://openaccess.thecvf.com/content_ECCV_2018/papers/Songtao_Liu_Receptive_Field_Block_ECCV_2018_paper.pdf
代碼鏈接：https://github.com/ruinmessi/RFBNet
實驗結果
26、PFP-NET
論文：Parallel Feature Pyramid Network for Object Detection
論文鏈接：
http://openaccess.thecvf.com/content_ECCV_2018/papers/Seung-Wook_Kim_Parallel_Feature_Pyramid_ECCV_2018_paper.pdf
代碼鏈接：
實驗結果
27、TS2C
論文：TS2C: Tight Box Mining with Surrounding Segmentation Context for Weakly Supervised Object Detection
論文鏈接：
http://openaccess.thecvf.com/content_ECCV_2018/papers/Yunchao_Wei_TS2C_Tight_Box_ECCV_2018_paper.pdf
代碼鏈接：
實驗結果
28、SAN
論文：
SAN: Learning Relationship between Convolutional Features for Multi-Scale Object Detection
論文鏈接：
http://openaccess.thecvf.com/content_ECCV_2018/papers/Kim_SAN_Learning_Relationship_ECCV_2018_paper.pdf
代碼鏈接：
實驗結果
29、
論文：Deep Regionlets for Object Detection
論文鏈接：
http://openaccess.thecvf.com/content_ECCV_2018/papers/Hongyu_Xu_Deep_Regionlets_for_ECCV_2018_paper.pdf
代碼鏈接：
實驗結果
30、
論文：Context Refinement for Object Detection
論文鏈接：
http://openaccess.thecvf.com/content_ECCV_2018/papers/Zhe_Chen_Context_Refinement_for_ECCV_2018_paper.pdf
代碼鏈接：
實驗結果
