honglinchu / siamtrackers Goto Github PK

Very good work!
how to add the wechat group?
Thank you!

请问学长这个文件在哪里可以下载呢，或者应该如何生成呢，麻烦告知一二，谢谢了！！！

在生成训练样本时，请问大神，iou设置为小于多少合适呢

学长，您好。readme里百度网盘的数据集，打不开，无法下载。可以重新分享下吗。UAV123 BaiduYun password: 2iq4

VOT2018 BaiduYun password: e5eh

十分感激，谢谢学长

师兄您好，在训练之前需要先裁剪好GOT10K数据集吗，如果直接用原始的数据集训练，是不是对cpu负担很大，导致最后GPU跑不起来

出现ZeroDivisionError: division by zero，看了一下主要是在ious = np.nanmean(ious, axis=0)这句，并且查看了ious的值。

ious= [[nan nan nan ... array([0.02229282]) array([0.10334186])
array([0.13032249])]]
Traceback (most recent call last):
File "/home/ubuntu/code/SiamTrackers-master/2-SiamFC/SiamFC-GOT/bin/my_test.py", line 72, in
prec_score,succ_score,succ_rate=e.report([tracker.name])
File "/home/ubuntu/code/SiamTrackers-master/2-SiamFC/SiamFC-GOT/got10k/experiments/vot.py", line 417, in report
ious = np.nanmean(ious, axis=0)

您好，请问您这里的百度云网盘TrackingNet数据集和官方的TrackingNet数据集内容是一样的不

如何测试OTB该如何训练，您是否有OTB的测试模型。

I am unable to train these trackers due to hardware constraints. Can someone please provide pre-trained weights. I specifically need for SiamFC++.

大佬您好，请问有没有Youtubebb原始数据集的百度云下载链接

您好，感谢您收集的代码，对我帮助很大。我运行了SiamFC-VID里面的代码，下面是一些问题，希望您能解答。

1. 我从百度云盘上下载了您提供的siamfc_30.pth预训练模型，当我运行 my_test.py后，程序可以运行，但是和您README.md里面提供的accuracy和precision完全不同，性能非常差，只有0.1多。
2. 当我运行可视化程序查看OTB100中每个视频的跟踪结果时，发现跟踪结果非常差，完全漂移了，并且bbox非常大。如下图所示：
   
3. 我运行SiamFC-GOT时，效果挺好的，基本和您提供的README.md一样

用alexnet训练后，在otb和vot2018上进行测试，然后改成resnet22训练，再在otb和vot2018上测试，得到的结果没有变，一个数字都没变，这是为什么。

大佬您好，我看到你的rpn++复现的结果也不太好，您后面有没有尝试复现出原效果

学长，有没有训练好的SiameseFC模型呢

您好，感谢分享：）
想问一下，自己拍了自行车的几个小时的视频（假设用siamrpn++，只跟踪mobike类别），是不是先用labelimg之类的工具对该跟踪目标进行打标签，之后的步骤怎么处理呢？
多谢！

您好，我又来问问题了，不好意思问题有点多。我按照您的README文件训练后发现越训练越差，不知道怎么回事。我在训练过程中，每个stage x-1学习率设置为lrs = np.array([[6,7],[7,8]])，每个stage x-2学习率设置为lrs = np.array([[8,9],[9,10]])。在进行下一阶段训练前都会加载上一阶段的最好模型生成新的训练数据，但是效果是越来越差，不知道怎么回事，感觉很迷惑。

您好，我用您提供的程序在ubuntu16.04下面生成训练数据时出现MemoryError问题。当我按照网上的说明 cat /proc/sys/vm/overcommit_memory改为1时，又显示已杀死，没法生成训练数据。请问您是否遇到过同样的问题。或者您是否能够把训练数据分享一下，谢谢。

请问是否有运行自己视频进行跟踪的代码，只有测试数据集的代码吗

wechat QR code out of date，I want to join you !

作者您好，我想问下DA是一种互相关操作吗？哪篇
论文中提出来的啊。

root_dir = os.path.abspath('/media/xiaofei/ECD67E80D67E4B3A/test_data/vot
2018') # VOT测试暂时遇到问题

林林大佬您好，GOT10K数据集中提供了visible ratios，请教下您，这个visible ratios是怎么计算得到的？论文中只看到这句相关的The regions that are occluded or truncated by image boundary correspond to the invisible parts，麻烦您给指导下？不胜感激

All the models are uploaded at the Baidu, Baidu is not accessible for us,

Can you upload the models and everything is necessary for running the codes on the Google Drive?

您好，我又发现一个问题。我用DaSiamRPN-GOT加载官方提供的SiamRPNBIG.model，测试OTB2015结果为：
prec_score:0.840 -succ_score:0.634 -succ_rate:0.824。用UpdateNet（选择的step=1）加载SiamRPNBIG.model测试OTB2015结果为：prec_score:0.853 -succ_score:0.641 -succ_rate:0.830。这两个结果不一样，而且低于SiamR-CNN论文中提供的DaSiamRPN评测结果（prec:0.878, succ:0.658）。请问您知道是怎么回事吗？在VOT2018上测试DaSiamRPN结果为acc: 0.586, eao:0.383这个结果还算正常。

Hi.
Can you please share some transfer learning strategy to start with
without training from scratch?
thank you

可以分享一下单目标跟踪里面关于参数调节的一些技巧吗

Traceback (most recent call last):
File "./bin/my_test.py", line 66, in
tester.test()
File "/home/c506/lrh/SiamFC++/SiamFCpp-master/siamfcpp/engine/tester/tester_impl/vot.py", line 93, in test
self.run_tracker()
File "/home/c506/lrh/SiamFC++/SiamFCpp-master/siamfcpp/engine/tester/tester_impl/vot.py", line 121, in run_tracker
self.worker(keys, all_devs[0], result_queue, speed_queue)
File "/home/c506/lrh/SiamFC++/SiamFCpp-master/siamfcpp/engine/tester/tester_impl/vot.py", line 174, in worker
lost, speed = self.track_single_video(tracker, video, v_id=v_id)
File "/home/c506/lrh/SiamFC++/SiamFCpp-master/siamfcpp/engine/tester/tester_impl/vot.py", line 238, in track_single_video
package="vot_overlap").vot_overlap
File "/home/c506/anaconda3/envs/siamfc/lib/python3.6/importlib/init.py", line 126, in import_module
return _bootstrap._gcd_import(name[level:], package, level)
File "", line 994, in _gcd_import
File "", line 971, in _find_and_load
File "", line 941, in _find_and_load_unlocked
File "", line 219, in _call_with_frames_removed
File "", line 994, in _gcd_import
File "", line 971, in _find_and_load
File "", line 955, in _find_and_load_unlocked
File "", line 665, in _load_unlocked
File "", line 678, in exec_module
File "", line 219, in _call_with_frames_removed
File "/home/c506/lrh/SiamFC++/SiamFCpp-master/siamfcpp/evaluation/vot_benchmark/pysot/utils/init.py", line 10, in
from . import region
ImportError: cannot import name 'region'
主要是不能加载SiamFC++/SiamFCpp-master/siamfcpp/evaluation/vot_benchmark/pysot/utils/init.py中的region模块
我用pycharm打开看了一下，好像是相对路径导入的问题，新手对这个问题不知道怎么解决，希望学长答复一下。谢谢

up主你好，pysot带的vot只能评测出eao的数据，不能画vot的评估图，对于vot评估图请问你有好的解决方法吗？

Thank you for sharing and i'd like to ask if 2020 AAAI's global track will be shared.
（GlobalTrack: A Simple and Strong Baseline for Long-Term Tracking）

同学，你用siamfcpp官方代码训练过吗？我用一片V100在got10k上训练，速度感人，一个epoch要10个小时。你写的代码训练速度是怎样的？

我在训练时，发现单卡可以正常训练，两张卡时就报错了，请问是跟哪些变量有关啊？

报错如下，求问如何解决：

[2021-05-30 22:39:23,961-rk0-my_train_LT.py#293] init done
fatal: Not a git repository (or any parent up to mount point /media/hdc/data4)
Stopping at filesystem boundary (GIT_DISCOVERY_ACROSS_FILESYSTEM not set).
fatal: Not a git repository (or any parent up to mount point /media/hdc/data4)
Stopping at filesystem boundary (GIT_DISCOVERY_ACROSS_FILESYSTEM not set).
[2021-05-30 22:39:23,992-rk0-my_train_LT.py#308] Version Information:
commit :
log :

[2021-05-30 22:39:23,994-rk0-my_train_LT.py#309] config
{
"META_ARC": "siamrpn_r50_l234_dwxcorr_lt",
"CUDA": true,
"TRAIN": {
"THR_HIGH": 0.6,
"THR_LOW": 0.3,
"NEG_NUM": 16,
"POS_NUM": 16,
"TOTAL_NUM": 64,
"EXEMPLAR_SIZE": 127,
"SEARCH_SIZE": 255,
"BASE_SIZE": 8,
"OUTPUT_SIZE": 25,
"RESUME": "",
"LOG_DIR": "./models/siamrpn_lt/logs",
"SNAPSHOT_DIR": "./models/siamrpn_lt/snapshot",
"EPOCH": 20,
"START_EPOCH": 0,
"BATCH_SIZE": 8,
"NUM_WORKERS": 1,
"MOMENTUM": 0.9,
"WEIGHT_DECAY": 0.0001,
"CLS_WEIGHT": 1.0,
"LOC_WEIGHT": 1.2,
"MASK_WEIGHT": 1,
"PRINT_FREQ": 20,
"LOG_GRADS": false,
"GRAD_CLIP": 10.0,
"BASE_LR": 0.005,
"LR": {
"TYPE": "log",
"KWARGS": {
"start_lr": 0.005,
"end_lr": 0.0005
}
},
"LR_WARMUP": {
"WARMUP": true,
"TYPE": "step",
"EPOCH": 5,
"KWARGS": {
"start_lr": 0.001,
"end_lr": 0.005,
"step": 1
}
}
},
"DATASET": {
"TEMPLATE": {
"SHIFT": 4,
"SCALE": 0.05,
"BLUR": 0.0,
"FLIP": 0.0,
"COLOR": 1.0
},
"SEARCH": {
"SHIFT": 64,
"SCALE": 0.18,
"BLUR": 0.2,
"FLIP": 0.0,
"COLOR": 1.0
},
"NEG": 0.2,
"GRAY": 0.0,
"NAMES": [
"GOT"
],
"VID": {
"ROOT": "data/crop511",
"ANNO": "data/crop511/train/train.json",
"FRAME_RANGE": 100,
"NUM_USE": 100000
},
"GOT": {
"ROOT": "./data/crop511",
"ANNO": "./data/crop511/train/train.json",
"FRAME_RANGE": 100,
"NUM_USE": 64000
},
"YOUTUBEBB": {
"ROOT": "data/yt_bb/crop511",
"ANNO": "data/yt_bb/train.json",
"FRAME_RANGE": 3,
"NUM_USE": -1
},
"COCO": {
"ROOT": "data/coco/crop511",
"ANNO": "data/coco/train2017.json",
"FRAME_RANGE": 1,
"NUM_USE": -1
},
"DET": {
"ROOT": "data/det/crop511",
"ANNO": "data/det/train.json",
"FRAME_RANGE": 1,
"NUM_USE": -1
},
"VIDEOS_PER_EPOCH": 64000
},
"BACKBONE": {
"TYPE": "resnet50",
"KWARGS": {
"used_layers": [
2,
3,
4
]
},
"PRETRAINED": "",
"TRAIN_LAYERS": [
"layer2",
"layer3",
"layer4"
],
"LAYERS_LR": 0.1,
"TRAIN_EPOCH": 10
},
"ADJUST": {
"ADJUST": true,
"KWARGS": {
"in_channels": [
512,
1024,
2048
],
"out_channels": [
128,
256,
512
]
},
"TYPE": "AdjustAllLayer"
},
"RPN": {
"TYPE": "MultiRPN",
"KWARGS": {
"anchor_num": 5,
"in_channels": [
128,
256,
512
],
"weighted": true
}
},
"MASK": {
"MASK": false,
"TYPE": "MaskCorr",
"KWARGS": {}
},
"REFINE": {
"REFINE": false,
"TYPE": "Refine"
},
"ANCHOR": {
"STRIDE": 8,
"RATIOS": [
0.33,
0.5,
1,
2,
3
],
"SCALES": [
8
],
"ANCHOR_NUM": 5
},
"TRACK": {
"TYPE": "SiamRPNLTTracker",
"PENALTY_K": 0.05,
"WINDOW_INFLUENCE": 0.28,
"LR": 0.22,
"EXEMPLAR_SIZE": 127,
"INSTANCE_SIZE": 255,
"BASE_SIZE": 8,
"CONTEXT_AMOUNT": 0.5,
"LOST_INSTANCE_SIZE": 831,
"CONFIDENCE_LOW": 0.8,
"CONFIDENCE_HIGH": 0.998,
"MASK_THERSHOLD": 0.3,
"MASK_OUTPUT_SIZE": 127
}
}
[2021-05-30 22:39:27,745-rk0-my_train_LT.py# 75] build train dataset
[2021-05-30 22:39:27,747-rk0-dataset.py# 38] loading GOT
[2021-05-30 22:39:27,759-rk0-dataset.py# 63] GOT loaded
[2021-05-30 22:39:27,783-rk0-dataset.py# 92] GOT start-index 0 select [64000/4] path_format {}.{}.{}.bmp
[2021-05-30 22:39:27,832-rk0-dataset.py#202] shuffle done!
[2021-05-30 22:39:27,833-rk0-dataset.py#203] dataset length 1280000
[2021-05-30 22:39:27,841-rk0-my_train_LT.py# 78] build dataset done
/home/hdc/anaconda3/envs/SiamTrackers/lib/python3.7/site-packages/torch/optim/lr_scheduler.py:122: UserWarning: Detected call of lr_scheduler.step() before optimizer.step(). In PyTorch 1.1.0 and later, you should call them in the opposite order: optimizer.step() before lr_scheduler.step(). Failure to do this will result in PyTorch skipping the first value of the learning rate schedule. See more details at https://pytorch.org/docs/stable/optim.html#how-to-adjust-learning-rate
"https://pytorch.org/docs/stable/optim.html#how-to-adjust-learning-rate", UserWarning)
[2021-05-30 22:39:27,844-rk0-my_train_LT.py#347] (WarmUPScheduler) lr spaces:
[0.001 0.00137973 0.00190365 0.00262653 0.0036239 0.005
0.00424171 0.00359843 0.0030527 0.00258974 0.00219699 0.0018638
0.00158114 0.00134135 0.00113792 0.00096535 0.00081895 0.00069475
0.00058938 0.0005 ]
[2021-05-30 22:39:27,852-rk0-my_train_LT.py#348] model prepare done
/home/hdc/anaconda3/envs/SiamTrackers/lib/python3.7/site-packages/torch/nn/parallel/_functions.py:61: UserWarning: Was asked to gather along dimension 0, but all input tensors were scalars; will instead unsqueeze and return a vector.
warnings.warn('Was asked to gather along dimension 0, but all '
[2021-05-30 22:39:34,181-rk0-my_train_LT.py#282] Epoch: [1][20/8000] lr: 0.001000
batch_time: 0.161238 (0.315044) data_time: 0.000092 (0.011209)
total_loss: 0.860676 (1.279064) cls_loss: 0.482578 (0.566196)
loc_loss: 0.315082 (0.594057)
[2021-05-30 22:39:34,182-rk0-log_helper.py#111] Progress:20/160000,[0%], Speed:0.315 s/iter,Epoch-period 42:0 (M:S),ETA 0:14:00(D:H:M)

[2021-05-30 22:39:37,413-rk0-my_train_LT.py#282] Epoch: [1][40/8000] lr: 0.001000
batch_time: 0.162611 (0.237624) data_time: 0.000102 (0.005661)
total_loss: 1.051070 (1.156314) cls_loss: 0.447658 (0.500955)
loc_loss: 0.502843 (0.546132)
[2021-05-30 22:39:37,413-rk0-log_helper.py#111] Progress:40/160000,[0%], Speed:0.238 s/iter,Epoch-period 31:40 (M:S),ETA 0:10:33(D:H:M)

Traceback (most recent call last):
File "/media/hdc/data4/wxl/SiamTrackers-master/7-SiamRPNpp/SiamRPNpp-DW/bin/my_train_LT.py", line 357, in
main()
File "/media/hdc/data4/wxl/SiamTrackers-master/7-SiamRPNpp/SiamRPNpp-DW/bin/my_train_LT.py", line 351, in main
train(train_loader, dist_model, optimizer, lr_scheduler, tb_writer)
File "/media/hdc/data4/wxl/SiamTrackers-master/7-SiamRPNpp/SiamRPNpp-DW/bin/my_train_LT.py", line 237, in train
outputs = model(data)
File "/home/hdc/anaconda3/envs/SiamTrackers/lib/python3.7/site-packages/torch/nn/modules/module.py", line 532, in call
result = self.forward(*input, **kwargs)
File "/home/hdc/anaconda3/envs/SiamTrackers/lib/python3.7/site-packages/torch/nn/parallel/data_parallel.py", line 152, in forward
outputs = self.parallel_apply(replicas, inputs, kwargs)
File "/home/hdc/anaconda3/envs/SiamTrackers/lib/python3.7/site-packages/torch/nn/parallel/data_parallel.py", line 162, in parallel_apply
return parallel_apply(replicas, inputs, kwargs, self.device_ids[:len(replicas)])
File "/home/hdc/anaconda3/envs/SiamTrackers/lib/python3.7/site-packages/torch/nn/parallel/parallel_apply.py", line 85, in parallel_apply
output.reraise()
File "/home/hdc/anaconda3/envs/SiamTrackers/lib/python3.7/site-packages/torch/_utils.py", line 394, in reraise
raise self.exc_type(msg)
RuntimeError: Caught RuntimeError in replica 1 on device 1.
Original Traceback (most recent call last):
File "/home/hdc/anaconda3/envs/SiamTrackers/lib/python3.7/site-packages/torch/nn/parallel/parallel_apply.py", line 60, in _worker
output = module(*input, **kwargs)
File "/home/hdc/anaconda3/envs/SiamTrackers/lib/python3.7/site-packages/torch/nn/modules/module.py", line 532, in call
result = self.forward(*input, **kwargs)
File "/media/hdc/data4/wxl/SiamTrackers-master/7-SiamRPNpp/SiamRPNpp-DW/siamrpnpp/models/model_builder.py", line 112, in forward
cls_loss = select_cross_entropy_loss(cls, label_cls)
File "/media/hdc/data4/wxl/SiamTrackers-master/7-SiamRPNpp/SiamRPNpp-DW/siamrpnpp/models/loss.py", line 25, in select_cross_entropy_loss
loss_pos = get_cls_loss(pred, label, pos)
File "/media/hdc/data4/wxl/SiamTrackers-master/7-SiamRPNpp/SiamRPNpp-DW/siamrpnpp/models/loss.py", line 17, in get_cls_loss
return F.nll_loss(pred, label)
File "/home/hdc/anaconda3/envs/SiamTrackers/lib/python3.7/site-packages/torch/nn/functional.py", line 1838, in nll_loss
ret = torch._C._nn.nll_loss(input, target, weight, _Reduction.get_enum(reduction), ignore_index)
RuntimeError: invalid argument 2: non-empty vector or matrix expected at /pytorch/aten/src/THCUNN/generic/ClassNLLCriterion.cu:31

ValueError: max() arg is an empty sequence
您好，我想问一下在测试的时候已经得到跟踪框的位置，但最后显示这个错误，没能得到eao

请问有人能帮忙解决一下这个问题吗

Traceback (most recent call last):
File "K:/SiamTrackers-master/2-SiamFC/SiamFC-GOT/bin/my_test.py", line 8, in
from got10k.experiments import *
File "K:\SiamTrackers-master\2-SiamFC\SiamFC-GOT\got10k\experiments\got10k.py", line 13, in
from ..datasets import GOT10k
ImportError: attempted relative import with no known parent package

我看前几天其实有人提问过这个问题，但是作者没有解答，我再提问一次。有人懂这个跨文件调用怎么出错了吗？应该1怎么改呀？

您好，我下载了部分GOT 数据，看能不能跑通代码，结果运行出来这样的结果，后面就结束了，没有报epoch loss 等，

请问这是哪里出现问题了，在代码里定位不到这一块

请问如何实现迁移到自己的数据集上呢？

attempted relative import with no known parent package

请问学长 这个问题怎么解决啊
在siamfc程序里出现的

感谢您的贡献！
我在使用过程中出现了一些问题。
我使用的是siamrpn，我在训练的时候出现了数据加载太慢的情况，我已经将batchsize改为1了，迟迟无法获得数据。

我在dataset下调用还是挺快的，train_dataset.getitem(1)， 但是不知道为什么在Dataloader中无法输出数据。

您好，我在训练alexnet的SiamRPNpp时，加载模型和数据集正常，但训练两次都出现“已杀死”的打印信息后就结束了训练，一次到99%，一次到80%，请问是什么问题

如题

https://github.com/HonglinChu/SiamTrackers/blob/master/7-SiamRPNpp/SiamRPNpp-DW/siamrpnpp/datasets/dataset.py
作者你好，这里的dataset文件似乎没有对输入的图片做从0-255到0-1的映射，按理说一般网络不应该会对输入的图像数据统一到0-1之间吗？

Why there are too many wrong annotated bbox for GOT10K dataset? For example, GOT-10k_Train_009058 sequence.

学长，我按照siamFCpp源码训练之后测试了一下，然后将测试得到的.zip文件放到got10官网上评测，却不出结果，格式完全跟他给的fc实例一样，这种情况你有了解，有什么解决办法吗？？

作者您好，我想问下使用got10K工具评估VOT数据集会报错吗？
我是报了一个除零错，就是在对iou除去NaN数据后做平均的时候，请问这个您解决了吗？