SiameseFC-pytorch-master
所属分类：其他
开发工具：Python
文件大小：9778KB
下载次数：13
上传日期：2019-03-16 16:04:22
上 传 者：
是否会对
说明： 基于孪生网络的目标跟踪，是现阶段比较流行的方法之一
(Target tracking based on twin network and target tracking based on twin network)
文件列表:
track (0, 2018-09-05)
track\2016-08-17.net.mat (8721150, 2018-09-05)
track\eval_otb.py (6465, 2018-09-05)
track\net.py (5342, 2018-09-05)
track\track_Siamfc.py (8907, 2018-09-05)
track\utils.py (4652, 2018-09-05)
train (0, 2018-09-05)
train\ILSVRC2015.stats.mat (745433, 2018-09-05)
train\dataset.py (11297, 2018-09-05)
train\dataset (0, 2018-09-05)
train\dataset\OTB2013.json (2625159, 2018-09-05)
train\dataset\OTB2015.json (1597279, 2018-09-05)
train\dataset\datase (1, 2018-09-05)
train\dataset\gen_otb2013.py (829, 2018-09-05)
train\eval_otb.py (6465, 2018-09-05)
train\loss.py (303, 2018-09-05)
train\net.py (5342, 2018-09-05)
train\train_Siamfc.py (22960, 2018-09-05)
train\utils.py (7448, 2018-09-05)
# SiameseFC-pytorch
pytorch version of the SiameseFC tracker descirbed in the paper "Fully-Convolutional Siamese nets for object tracking".
This repository includes training and tracking codes.
Note that
(1) The tracking results using the official model (2016-08-17.net.mat) are 2% worse than the original matlab code (https://github.com/bertinetto/siamese-fc).
(2) The training codes exist bugs that cause poor performance (nearly 4% worse than the offical report).
We are looking for these bugs, and hope you can join us! Furthermore, these codes are added comments to easy understanding.
Finally, we thank all the authors in github. If you have any questions/ are interested in object tracking, please feel free to contact me.
近期下载者：
相关文件：
收藏者：