原代码地址：

GitHub - VISION-SJTU/IoUattack: [CVPR2021] IoU Attack: Towards Temporally Coherent Black-Box Adversarial Attack for Visual Object Tracking

1.test_original

数据集报错处理：

每个类设置为，如siamrpn_r50_l234_dwxcorr/ants1/color/00000001.jpg

no module named pysot:

官方方法：

export PYTHONPATH=path/to/pysot

sys方法：

将pysot所在的父文件夹（pysot-master）,用sys进行导入，即

import sys

sys.path.append('/media/blingbling/store/code/pysot-master')

终端输入语句：

cd ../experiments/siamrpn_r50_l234_dwxcorr

python -u ../../tools/test_original.py --snapshot model.pth --dataset VOT2018 --config config.yaml

ValueError: operands could not be broadcast together with shapes (3125,) (2205,):

原因：修改了config文件中的batch size,只修改了一个，导致不匹配，改回原来的数字即可

输出结果：

( 1) Video: ants1 Time: 70.6s Speed: 4.6fps Lost: 0
( 2) Video: ants3 Time: 136.0s Speed: 4.3fps Lost: 1
( 3) Video: bag Time: 47.1s Speed: 4.1fps Lost: 0
( 4) Video: ball1 Time: 24.3s Speed: 4.3fps Lost: 0
( 5) Video: ball2 Time: 9.4s Speed: 4.3fps Lost: 0
( 6) Video: basketball Time: 168.1s Speed: 4.3fps Lost: 1
( 7) Video: birds1 Time: 79.1s Speed: 4.3fps Lost: 0
( 8) Video: blanket Time: 51.4s Speed: 4.4fps Lost: 1
( 9) Video: bmx Time: 17.7s Speed: 4.2fps Lost: 0
( 10) Video: bolt1 Time: 81.6s Speed: 4.3fps Lost: 0
( 11) Video: bolt2 Time: 66.5s Speed: 4.4fps Lost: 1
( 12) Video: book Time: 39.2s Speed: 4.4fps Lost: 1
( 13) Video: butterfly Time: 34.8s Speed: 4.3fps Lost: 0
( 14) Video: car1 Time: 171.8s Speed: 4.3fps Lost: 0
( 15) Video: conduction1 Time: 82.9s Speed: 4.1fps Lost: 0
( 16) Video: crabs1 Time: 35.2s Speed: 4.5fps Lost: 3
( 17) Video: crossing Time: 29.5s Speed: 4.4fps Lost: 0
( 18) Video: dinosaur Time: 90.8s Speed: 3.6fps Lost: 2
( 19) Video: drone1 Time: 83.0s Speed: 4.3fps Lost: 3
( 20) Video: drone_across Time: 36.6s Speed: 4.0fps Lost: 0
( 21) Video: drone_flip Time: 24.4s Speed: 4.6fps Lost: 2
( 22) Video: fernando Time: 67.8s Speed: 4.3fps Lost: 2
( 23) Video: fish1 Time: 91.7s Speed: 4.0fps Lost: 1
( 24) Video: fish2 Time: 72.6s Speed: 4.3fps Lost: 1
( 25) Video: fish3 Time: 123.8s Speed: 4.2fps Lost: 0
( 26) Video: flamingo1 Time: 322.0s Speed: 4.3fps Lost: 2
( 27) Video: frisbee Time: 57.2s Speed: 4.3fps Lost: 1
( 28) Video: girl Time: 351.5s Speed: 4.3fps Lost: 2
( 29) Video: glove Time: 27.0s Speed: 4.4fps Lost: 1
( 30) Video: godfather Time: 84.9s Speed: 4.3fps Lost: 1
( 31) Video: graduate Time: 200.3s Speed: 4.2fps Lost: 0
( 32) Video: gymnastics1 Time: 131.2s Speed: 4.3fps Lost: 0
( 33) Video: gymnastics2 Time: 55.4s Speed: 4.3fps Lost: 0
( 34) Video: gymnastics3 Time: 26.1s Speed: 4.5fps Lost: 1
( 35) Video: hand Time: 60.6s Speed: 4.4fps Lost: 1
( 36) Video: handball1 Time: 85.0s Speed: 4.4fps Lost: 2
( 37) Video: handball2 Time: 92.4s Speed: 4.3fps Lost: 3
( 38) Video: helicopter Time: 155.0s Speed: 4.6fps Lost: 1
( 39) Video: iceskater1 Time: 150.6s Speed: 4.4fps Lost: 0
( 40) Video: iceskater2 Time: 156.9s Speed: 4.5fps Lost: 1
( 41) Video: leaves Time: 12.1s Speed: 5.1fps Lost: 2
( 42) Video: matrix Time: 21.8s Speed: 4.5fps Lost: 0
( 43) Video: motocross1 Time: 35.0s Speed: 4.7fps Lost: 1
( 44) Video: motocross2 Time: 13.2s Speed: 4.6fps Lost: 0
( 45) Video: nature Time: 219.8s Speed: 4.5fps Lost: 3
( 46) Video: pedestrian1 Time: 30.4s Speed: 4.6fps Lost: 0
( 47) Video: rabbit Time: 33.4s Speed: 4.7fps Lost: 3
( 48) Video: racing Time: 34.9s Speed: 4.4fps Lost: 0
( 49) Video: road Time: 120.4s Speed: 4.6fps Lost: 0
( 50) Video: shaking Time: 78.6s Speed: 4.6fps Lost: 0
( 51) Video: sheep Time: 54.0s Speed: 4.6fps Lost: 0
( 52) Video: singer2 Time: 79.1s Speed: 4.6fps Lost: 0
( 53) Video: singer3 Time: 28.2s Speed: 4.6fps Lost: 0
( 54) Video: soccer1 Time: 82.4s Speed: 4.7fps Lost: 3
( 55) Video: soccer2 Time: 26.7s Speed: 4.8fps Lost: 1
( 56) Video: soldier Time: 28.8s Speed: 4.8fps Lost: 1
( 57) Video: tiger Time: 77.9s Speed: 4.7fps Lost: 0
( 58) Video: traffic Time: 42.4s Speed: 4.5fps Lost: 0
( 59) Video: wiper Time: 71.6s Speed: 4.7fps Lost: 1
( 60) Video: zebrafish1 Time: 85.3s Speed: 4.7fps Lost: 0
SiamRPN++(Original) total lost: 50

共丢50个，每类最多丢3个；

2.test_IoU attack

( 1) Video: ants1 Time: 1991.0s Speed: 0.2fps Lost: 11
( 2) Video: ants3 Time: 4093.5s Speed: 0.1fps Lost: 14
( 3) Video: bag Time: 297.1s Speed: 0.7fps Lost: 0
( 4) Video: ball1 Time: 806.6s Speed: 0.1fps Lost: 5
( 5) Video: ball2 Time: 371.9s Speed: 0.1fps Lost: 1
( 6) Video: basketball Time: 1740.3s Speed: 0.4fps Lost: 4
( 7) Video: birds1 Time: 2429.5s Speed: 0.1fps Lost: 15
( 8) Video: blanket Time: 518.2s Speed: 0.4fps Lost: 0
( 9) Video: bmx Time: 519.0s Speed: 0.1fps Lost: 0
( 10) Video: bolt1 Time: 2067.8s Speed: 0.2fps Lost: 1
( 11) Video: bolt2 Time: 1007.9s Speed: 0.3fps Lost: 3
( 12) Video: book Time: 1447.2s Speed: 0.1fps Lost: 7
( 13) Video: butterfly Time: 1338.7s Speed: 0.1fps Lost: 1
( 14) Video: car1 Time: 2985.2s Speed: 0.2fps Lost: 1
( 15) Video: conduction1 Time: 1843.3s Speed: 0.2fps Lost: 2
( 16) Video: crabs1 Time: 883.9s Speed: 0.2fps Lost: 6
( 17) Video: crossing Time: 297.2s Speed: 0.4fps Lost: 0
( 18) Video: dinosaur Time: 2375.0s Speed: 0.1fps Lost: 4
( 19) Video: drone1 Time: 2950.2s Speed: 0.1fps Lost: 12
( 20) Video: drone_across Time: 1183.5s Speed: 0.1fps Lost: 2
( 21) Video: drone_flip Time: 813.9s Speed: 0.1fps Lost: 4
( 22) Video: fernando Time: 1579.9s Speed: 0.2fps Lost: 3
( 23) Video: fish1 Time: 1277.3s Speed: 0.3fps Lost: 2
( 24) Video: fish2 Time: 2047.2s Speed: 0.2fps Lost: 3
( 25) Video: fish3 Time: 2010.5s Speed: 0.3fps Lost: 0
( 26) Video: flamingo1 Time: 11976.6s Speed: 0.1fps Lost: 2
( 27) Video: frisbee Time: 2531.8s Speed: 0.1fps Lost: 5
( 28) Video: girl Time: 4417.5s Speed: 0.3fps Lost: 4
( 29) Video: glove Time: 469.2s Speed: 0.3fps Lost: 3
( 30) Video: godfather Time: 1152.9s Speed: 0.3fps Lost: 2
( 31) Video: graduate Time: 3731.1s Speed: 0.2fps Lost: 5
( 32) Video: gymnastics1 Time: 3416.4s Speed: 0.2fps Lost: 2
( 33) Video: gymnastics2 Time: 884.7s Speed: 0.3fps Lost: 1
( 34) Video: gymnastics3 Time: 728.1s Speed: 0.2fps Lost: 2
( 35) Video: hand Time: 1944.3s Speed: 0.1fps Lost: 12
( 36) Video: handball1 Time: 1598.9s Speed: 0.2fps Lost: 6
( 37) Video: handball2 Time: 1729.0s Speed: 0.2fps Lost: 7
( 38) Video: helicopter Time: 2288.9s Speed: 0.3fps Lost: 2
( 39) Video: iceskater1 Time: 4523.5s Speed: 0.1fps Lost: 0
( 40) Video: iceskater2 Time: 4223.2s Speed: 0.2fps Lost: 2
( 41) Video: leaves Time: 380.0s Speed: 0.2fps Lost: 4
( 42) Video: matrix Time: 529.2s Speed: 0.2fps Lost: 3
( 43) Video: motocross1 Time: 1161.4s Speed: 0.1fps Lost: 1
( 44) Video: motocross2 Time: 274.3s Speed: 0.2fps Lost: 0
( 45) Video: nature Time: 2568.5s Speed: 0.4fps Lost: 3
( 46) Video: pedestrian1 Time: 1126.5s Speed: 0.1fps Lost: 1
( 47) Video: rabbit Time: 727.0s Speed: 0.2fps Lost: 8
( 48) Video: racing Time: 181.0s Speed: 0.9fps Lost: 0
( 49) Video: road Time: 1769.4s Speed: 0.3fps Lost: 1
( 50) Video: shaking Time: 921.1s Speed: 0.4fps Lost: 0
( 51) Video: sheep Time: 532.9s Speed: 0.5fps Lost: 0
( 52) Video: singer2 Time: 662.4s Speed: 0.6fps Lost: 0
( 53) Video: singer3 Time: 325.0s Speed: 0.4fps Lost: 2
( 54) Video: soccer1 Time: 2030.0s Speed: 0.2fps Lost: 5
( 55) Video: soccer2 Time: 957.6s Speed: 0.1fps Lost: 10
( 56) Video: soldier Time: 267.6s Speed: 0.5fps Lost: 2
( 57) Video: tiger Time: 1730.7s Speed: 0.2fps Lost: 2
( 58) Video: traffic Time: 632.7s Speed: 0.3fps Lost: 0
( 59) Video: wiper Time: 984.2s Speed: 0.3fps Lost: 2
( 60) Video: zebrafish1 Time: 2535.9s Speed: 0.2fps Lost: 4
SiamRPN++(IoU_attack) total lost: 204

共丢204个，最多丢15个。

IoUattack复现相关推荐

Facebook 发布深度学习工具包 PyTorch Hub，让论文复现变得更容易
近日,PyTorch 社区发布了一个深度学习工具包 PyTorchHub, 帮助机器学习工作者更快实现重要论文的复现工作.PyTorchHub 由一个预训练模型仓库组成,专门用于提高研究工作的复现性以 ...
【目标检测】yolo系列：从yolov1到yolov5之YOLOv1详解及复现
检测器通常能够被分为两类,一类是two-stage检测器,最具代表的为faster R-CNN:另一类是one-stage检测器,包括YOLO,SSD等.一般来说,two-stage检测器具有高定位和 ...
CV算法复现（分类算法2/6）：AlexNet（2012年 Hinton组）
致谢:霹雳吧啦Wz:https://space.bilibili.com/18161609 目录致谢:霹雳吧啦Wz:https://space.bilibili.com/18161609 1 本次要 ...
【神经网络】(17) EfficientNet 代码复现，网络解析，附Tensorflow完整代码
各位同学好,今天和大家分享一下如何使用 Tensorflow 复现 EfficientNet 卷积神经网络模型. EfficientNet 的网络结构和 MobileNetV3 比较相似,建议大家在学 ...
【神经网络】(16) MobileNetV3 代码复现，网络解析，附Tensorflow完整代码
各位同学好,今天和大家分享一下如何使用 Tensorflow 构建 MobileNetV3 轻量化网络模型. MobileNetV3 做了如下改动(1)更新了V2中的逆转残差结构:(2)使用NAS搜索 ...
【神经网络】(15) Xception 代码复现，网络解析，附Tensorflow完整代码
各位同学好,今天和大家分享一下如何使用 Tensorflow 构建 Xception 神经网络模型. 在前面章节中,我已经介绍了很多种轻量化卷积神经网络模型,感兴趣的可以看一下:https://blo ...
【神经网络】(14) MnasNet 代码复现，网络解析，附Tensorflow完整代码
各位同学好,今天和大家分享一下如何使用 Tensorflow 复现谷歌轻量化神经网络 MnasNet 通常而言,移动端(手机)和终端(安防监控.无人驾驶)上的设备计算能力有限,无法搭载庞大的神经网络 ...
【神经网络】(13) ShuffleNetV2 代码复现，网络解析，附Tensorflow完整代码
各位同学好,今天和大家分享一下如何使用 Tensorflow 复现轻量化神经网络 ShuffleNetV2. 为了能将神经网络模型用于移动端(手机)和终端(安防监控.无人驾驶)的实时计算,通常这些设备 ...
【神经网络】(11) 轻量化网络MobileNetV1代码复现、解析，附Tensorflow完整代码
各位同学好,今天和大家分享一下如何使用 Tensorflow 复现轻量化神经网络模型 MobileNetV1.为了能将神经网络模型用于移动端(手机)和终端(安防监控.无人驾驶)的实时计算,通常这些设备 ...

IoUattack复现