daikenan / ltmu Goto Github PK

I noticed that there is no evaluator for VOT-LT in this codebase. Can you provide some evaluator toolkits about VOT-LT?

Great work! May I ask if you could provide any code snippet for testing on one's own dataset?

For example:
img[min_y_val:max_y_val, min_x_val:max_x_val, :].shape = (3,3,3)
min_x = 1280, min_y = 673
max_x = 1283, max_y = 676
min_x_val = 1276, min_y_val = 673
max_x_val = 1279, max_y_val = 676
cropped.shape = (3,3,3)
cropped[min_y_val - min_y:max_y_val - min_y, min_x_val - min_x:max_x_val - min_x, :] = croppd[0:3,-4:-1,:],
its shape is (3,2,3)

hi, how can the method combined into siamese tracker? can you give me some advice?

thanks for your great job. When I run python run_tracker.py dimp dimp50 under the folder DiMP_LTMU/pytracking

I got a mistake: 'tuple' object has no attribute 'keys'

it seems that the the result of
def track function in tracker/dimp/dimp.py does not match the code in
tracker/base/basetracker.py

Maybe I used the wrong settings.

line 8 in the following block:
def forward_samples(model, image, samples, opts, out_layer='conv3'):
model.eval()
extractor = RegionExtractor(image, samples, opts)
for i, regions in enumerate(extractor):
if opts['use_gpu']:
regions = regions.cuda()
with torch.no_grad():
feat = model(regions, out_layer=out_layer)
if i==0:
feats = feat.detach().clone()
else:
feats = torch.cat((feats, feat.detach().clone()), 0)
return feats

should the 'feat' in feat = model(regions, out_layer=out_layer) be 'feats' ?

In the citation of DiMP_LTMU, [3] should be mdnet? I'm a little confused about you cite DIMP for [3].

hi kenan:
whe will you plan to release the code?

Hi I am wondering why your implementation of super dimp achieves 64.6 AUC on LaSOT which is 1.5 more than the official results? What are you doing differently? Thanks.

Hi.

Great work and solid results! I am happy to see that you are using our PyTracking framework. I hope you like it, and we appreciate any comments and suggestions :)

Since you base on our PyTracking framework, I would very much appreciate if you can mention this and link to our PyTracking in your readme.

Also, I saw in your code that you have separate instances of the PyTracking code in the DIMP_MU and DIMP_LTMU folders. Our framework has a quite general support of integrating several trackers into the PyTracking framework. So please send me a message if you want tips of how to integrate your trackers, or if you have any feedback to us how we can more easily support trackers like yours.

Best regards,
Martin

Hi, all the line work until the demo.py for which I got the following error:

Traceback (most recent call last):
File "demo.py", line 9, in
eval_tracking('demo', p=p)
File "/home/hexin/LTMU/DiMP_LTMU/run_tracker.py", line 187, in eval_tracking
run_seq_list(Dataset, p, sequence_list, data_dir)
File "/home/hexin/LTMU/DiMP_LTMU/run_tracker.py", line 138, in run_seq_list
tracker = Dimp_LTMU_Tracker(image, region, p=p, groundtruth=groundtruth)
File "/home/hexin/LTMU/DiMP_LTMU/Dimp_LTMU.py", line 75, in init
self.local_init(image, init_gt1)
File "/home/hexin/LTMU/DiMP_LTMU/Dimp_LTMU.py", line 370, in local_init
self.local_Tracker.initialize(image, init_box)
File "/home/hexin/LTMU/DiMP_LTMU/pytracking/tracker/dimp/dimp.py", line 27, in initialize
self.initialize_features()
File "/home/hexin/LTMU/DiMP_LTMU/pytracking/tracker/dimp/dimp.py", line 17, in initialize_features
self.params.net.initialize()
File "/home/hexin/LTMU/DiMP_LTMU/pytracking/features/net_wrappers.py", line 41, in initialize
super().initialize()
File "/home/hexin/LTMU/DiMP_LTMU/pytracking/features/net_wrappers.py", line 34, in initialize
self.load_network()
File "/home/hexin/LTMU/DiMP_LTMU/pytracking/features/net_wrappers.py", line 28, in load_network
self.net = load_network(self.net_path)
File "/home/hexin/LTMU/DiMP_LTMU/pytracking/utils/loading.py", line 23, in load_network
net, _ = ltr_loading.load_network(path_full, backbone_pretrained=False)
File "./ltr/admin/loading.py", line 44, in load_network
raise Exception('No matching checkpoint file found')
Exception: No matching checkpoint file found

Do you have any idea for the cause of this? Many thanks

Hi,
Can you publish your repository license?
Thanks!

Hi, thanks for sharing your code. I’m trying to run the demo script to test the DiMP_LTMU tracker but this error occurs:

Downloading: "https://download.pytorch.org/models/resnet50-19c8e357.pth" to /TBData2/kristian/.cache/torch/checkpoints/resnet50-19c8e357.pth
100%|██████████████████████████████████████| 97.8M/97.8M [00:01<00:00, 69.9MB/s]
0: yamaha: 1 /3143
1.36641526222229
Traceback (most recent call last):
  File "demo.py", line 9, in <module>
    eval_tracking('demo', p=p)
  File "/media/TBData2/kristian_projects/conda_env/LTMU_demo/DiMP_LTMU/run_tracker.py", line 187, in eval_tracking
    run_seq_list(Dataset, p, sequence_list, data_dir)
  File "/media/TBData2/kristian_projects/conda_env/LTMU_demo/DiMP_LTMU/run_tracker.py", line 156, in run_seq_list
    region, score_map, iou, score_max, dis = tracker.tracking(image)
  File "/media/TBData2/kristian_projects/conda_env/LTMU_demo/DiMP_LTMU/Dimp_LTMU.py", line 526, in tracking
    frame_id=self.i, score=md_score, mask=mask)
TypeError: show_res() got an unexpected keyword argument 'score'

Has anyone encountered the same error or knows what the problem is? Thank you so much

Hi.
I tried to make an environment to use PrDiMP_MU. I run below command and get an error.
pip install -r requirements.txt
ERROR Cound not find a version that satisfies the requirement mkl (from versions: none)

I think it is related to the jetson CPU being made by arm. Do you know how to solve it?
Thanks!

Thank you for your work. The link of https://drive.google.com/open?id=1o-btxlWWA6GlbwMGCGkzn2vAw9qv8D2z is not available yet. Are there any alternative links？

The above error occured when i run python demo.py. so whta's the problem and how to solve it.
ths for your reply.

hello, I not have git for some reason. Previously, I used repo by downloading zip package.
but, the repo need to install sub-module, can't download zip package.
do you have any solution?
Thank you!

Dear Author:
Thanks for publishing the source code of Meta updater.
After referring to LTR and to read me in AtomMU Iam still confused about the dataset shape and the labels annotation needed.
for testing I will train the superdimp on single class from the Lasot dataset.
after downloading Lasot dataset the data is separated in multiple files.
should I combine the data in one file? and should I change the annotation of the Lasot dataset?

Hi, while testing on my own dataset. I got the following error:
File "demo.py", line 9, in
eval_tracking('lasot', p=p)
File "/home/hexin/LTMU/DiMP_LTMU/run_tracker.py", line 184, in eval_tracking
run_seq_list(Dataset, p, sequence_list, data_dir)
File "/home/hexin/LTMU/DiMP_LTMU/run_tracker.py", line 138, in run_seq_list
tracker = Dimp_LTMU_Tracker(image, region, p=p, groundtruth=groundtruth)
File "/home/hexin/LTMU/DiMP_LTMU/Dimp_LTMU.py", line 74, in init
self.init_pymdnet(image, init_gt1)
File "/home/hexin/LTMU/DiMP_LTMU/Dimp_LTMU.py", line 192, in init_pymdnet
pos_feats = forward_samples(self.pymodel, image, pos_examples, opts)
File "/home/hexin/LTMU/DiMP_LTMU/pyMDNet/tracking/run_tracker.py", line 41, in forward_samples
return feats

May I ask for any suggestion?

Hi all,
Is there any solution for test the trained network on a video file?

Detected CUDA files, patching ldflags Emitting ninja build file /tmp/torch_extensions/_prroi_pooling/build.ninja... Building extension module _prroi_pooling... 1.5.1 Loading extension module _prroi_pooling... Traceback (most recent call last): File "./ltr/external/PreciseRoIPooling/pytorch/prroi_pool/functional.py", line 33, in _import_prroi_pooling verbose=True File "/home/yu/.conda/envs/LTMU/lib/python3.7/site-packages/torch/utils/cpp_extension.py", line 680, in load is_python_module) File "/home/yu/.conda/envs/LTMU/lib/python3.7/site-packages/torch/utils/cpp_extension.py", line 877, in _jit_compile return _import_module_from_library(name, build_directory, is_python_module) File "/home/yu/.conda/envs/LTMU/lib/python3.7/site-packages/torch/utils/cpp_extension.py", line 1084, in _import_module_from_library file, path, description = imp.find_module(module_name, [path]) File "/home/yu/.conda/envs/LTMU/lib/python3.7/imp.py", line 296, in find_module raise ImportError(_ERR_MSG.format(name), name=name) ImportError: No module named '_prroi_pooling'

Great work!
Could you please tell me which version of Ninja should be used ?

What should i do?

Thanks for your amazing work.

I encountered an error when executing git submodule init :

fatal: no submodule mapping found in .gitmodules for path 'PrDiMP_MU/ltr/external/PreciseRoIPooling'

Do you have any idea about this error?

I noticed that most of the examples in the repo are based on CF trackers. Most Siam based trackers don't have a update algorithm, so I think they do need the update mechanism. If you have any plan to implement LTMU on one Siam Tracker and update this repo, that will be great. Thanks!

I saw the results on oxuva data set in the paper. I would like to ask if there is any test code for this piece.
Thanks!