Official Implementation of Towards Sequence-Level Training for Visual Tracking (ECCV 2022)
License: GNU General Public License v3.0
I'm Minji Kim, a PhD student at the Computer Vision Lab at Seoul National University.
Great work!
When I run LaSOT database by 'python pytracking/run_tracker.py slt_transt slt_transt --dataset_name lasot, there is an error, says "No constructor for the given network. " And I look up the code, I find in ''ltr/admin/loading.py ", when loading the checkpoint "slt_transt", there is no 'constructor' in 'checkpoint_dict' . So is the given model "slt_transt" not right?
Thanks for replying.
I meet the problem when run "python setup.py build_ext --inplace" under ${SLTtrack_ROOT}/pysot_toolkit. The problem is reported as below:
toolkit/utils/region.pyx:185:19: cimported module has no attribute 'compute_polygon_overlap'
Traceback (most recent call last):
File "/mnt/zihan/SLTtrack-master/pysot_toolkit/setup.py", line 21, in
ext_modules=cythonize(ext_modules)
File "/home/zihansr/anaconda3/envs/slt/lib/python3.9/site-packages/Cython/Build/Dependencies.py", line 1134, in cythonize
cythonize_one(*args)
File "/home/zihansr/anaconda3/envs/slt/lib/python3.9/site-packages/Cython/Build/Dependencies.py", line 1301, in cythonize_one
raise CompileError(None, pyx_file)
Cython.Compiler.Errors.CompileError: toolkit/utils/region.pyx
I followed all the previous steps and installed all package well. Could you tell me how to solve this or where can I find the solutions? Thanks!
Great jobs! Thanks for you sharing. How long will it take to train a model, like TransT?
I'm confused about why you using the tracker and self.net at the same time, and in my opinion, self.net is enough for the whole actors, can you explain for that?
Hi,
I am attempting to train on either the TransT or TrDiMP models but I keep getting this error.
`
Using CuDNN Benchmark
Training: slt_transt slt_transt
WARNING: You are using tensorboardX instead sis you have a too old pytorch version.
number of params: 23016006
No matching checkpoint file found
Backend TkAgg is interactive backend. Turning interactive mode on.
Training crashed at epoch 1
Traceback for the error!
Traceback (most recent call last):
File "/home/usr/Classes/RL/SLTtrack/ltr/../ltr/trainers/base_trainer.py", line 70, in train
self.train_epoch()
File "/home/usr/Classes/RL/SLTtrack/ltr/../ltr/trainers/slt_transt_trainer.py", line 150, in train_epoch
self.cycle_dataset(loader)
File "/home/usr/Classes/RL/SLTtrack/ltr/../ltr/trainers/slt_transt_trainer.py", line 62, in cycle_dataset
explore_result = self.actor.explore(data)
File "/home/usr/Classes/RL/SLTtrack/ltr/../ltr/actors/slt_transt_actor.py", line 82, in explore
outputs = tracker.batch_track(here_images, here_gt_bbox, action_mode='half')
File "/home/usr/Classes/RL/SLTtrack/ltr/../pytracking/tracker/slt_transt/slt_transt.py", line 143, in batch_track
outputs = self.net.track_batch(x_crop)
File "/home/usr/Classes/RL/SLTtrack/ltr/../ltr/models/tracking/transt.py", line 115, in track_batch
hs = self.featurefusion_network(self.input_proj(src_template), mask_template, self.input_proj(src_search), mask_search, self.pos_template[-1], pos_search[-1])
File "/home/usr/miniconda3/envs/slt/lib/python3.9/site-packages/torch/nn/modules/module.py", line 1102, in _call_impl
return forward_call(*input, **kwargs)
File "/home/usr/Classes/RL/SLTtrack/ltr/../ltr/models/neck/featurefusion_network.py", line 53, in forward
hs = self.decoder(memory_search, memory_temp,
File "/home/usr/miniconda3/envs/slt/lib/python3.9/site-packages/torch/nn/modules/module.py", line 1102, in _call_impl
return forward_call(*input, **kwargs)
File "/home/usr/Classes/RL/SLTtrack/ltr/../ltr/models/neck/featurefusion_network.py", line 77, in forward
output = layer(output, memory, tgt_mask=tgt_mask,
File "/home/usr/miniconda3/envs/slt/lib/python3.9/site-packages/torch/nn/modules/module.py", line 1102, in _call_impl
return forward_call(*input, **kwargs)
File "/home/usr/Classes/RL/SLTtrack/ltr/../ltr/models/neck/featurefusion_network.py", line 165, in forward
return self.forward_post(tgt, memory, tgt_mask, memory_mask,
File "/home/usr/Classes/RL/SLTtrack/ltr/../ltr/models/neck/featurefusion_network.py", line 151, in forward_post
tgt2 = self.linear2(self.dropout(self.activation(self.linear1(tgt))))
File "/home/usr/miniconda3/envs/slt/lib/python3.9/site-packages/torch/nn/modules/module.py", line 1102, in _call_impl
return forward_call(*input, **kwargs)
File "/home/usr/miniconda3/envs/slt/lib/python3.9/site-packages/torch/nn/modules/linear.py", line 103, in forward
return F.linear(input, self.weight, self.bias)
File "/home/usr/miniconda3/envs/slt/lib/python3.9/site-packages/torch/nn/functional.py", line 1848, in linear
return torch._C._nn.linear(input, weight, bias)
RuntimeError: CUDA out of memory. Tried to allocate 128.00 MiB (GPU 0; 15.72 GiB total capacity; 11.81 GiB already allocated; 66.81 MiB free; 11.90 GiB reserved in total by PyTorch) If reserved memory is >> allocated memory try setting max_split_size_mb to avoid fragmentation. See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CONF
`
I have set PYTORCH_CUDA_ALLOC_CONF=max_split_size_mb:512 to try to manage memory as suggested, but I get the same error. I am running on Ubuntu 22.04, and using an NVIDIA RTX 3080 Ti which has 16GB RAM. I haven't currently found any, but was wondering if there are any hyperparameters I can change in order to allow training to happen (however slow that may be).
It is very pleasant working. When I tried to reproduce this work, I encountered this problem: in the RL training stage, the reward and IoU declined and were almost all negative. Hope the author gives me a favor, thanks a lot!!!
There are some of the training logs on TransT, using frame-level pretrained TransT.pth(don't change any parameters!),
[train: 120, 1 / 125] FPS: 5.7 (5.7) , Loss/total: 0.46932 , Loss/sl_clf: -0.00489 , Loss/bbox: 0.05105 , Loss/giou: 0.14373 , bbox_iou: 0.85878 , grad_norm: 38.99021 , reward: -0.00151 , e_mIoU: 0.79518 , mIoU: 0.79670 , mIoU10: 0.85185 , mIoU100: 0.84046
[train: 120, 2 / 125] FPS: 9.9 (35.9) , Loss/total: 0.81592 , Loss/sl_clf: 0.02264 , Loss/bbox: 0.04261 , Loss/giou: 0.13162 , bbox_iou: 0.87164 , grad_norm: 49.15634 , reward: 0.00632 , e_mIoU: 0.83388 , mIoU: 0.82755 , mIoU10: 0.85434 , mIoU100: 0.84025
[train: 120, 3 / 125] FPS: 12.0 (21.3) , Loss/total: 0.65003 , Loss/sl_clf: 0.01190 , Loss/bbox: 0.04164 , Loss/giou: 0.13170 , bbox_iou: 0.87135 , grad_norm: 45.62053 , reward: 0.00315 , e_mIoU: 0.84058 , mIoU: 0.83742 , mIoU10: 0.85545 , mIoU100: 0.84016
[train: 120, 4 / 125] FPS: 14.0 (28.1) , Loss/total: 0.58626 , Loss/sl_clf: 0.00853 , Loss/bbox: 0.04010 , Loss/giou: 0.12892 , bbox_iou: 0.87407 , grad_norm: 40.98493 , reward: 0.00227 , e_mIoU: 0.84904 , mIoU: 0.84677 , mIoU10: 0.85537 , mIoU100: 0.84013
[train: 120, 5 / 125] FPS: 15.7 (31.0) , Loss/total: 0.55694 , Loss/sl_clf: 0.00660 , Loss/bbox: 0.03938 , Loss/giou: 0.13053 , bbox_iou: 0.87255 , grad_norm: 46.99872 , reward: 0.00167 , e_mIoU: 0.85015 , mIoU: 0.84848 , mIoU10: 0.85697 , mIoU100: 0.84000
[train: 120, 6 / 125] FPS: 15.8 (15.9) , Loss/total: 0.51575 , Loss/sl_clf: 0.00556 , Loss/bbox: 0.03696 , Loss/giou: 0.12379 , bbox_iou: 0.87896 , grad_norm: 44.13844 , reward: 0.00139 , e_mIoU: 0.85944 , mIoU: 0.85804 , mIoU10: 0.85827 , mIoU100: 0.84021
[train: 120, 7 / 125] FPS: 16.5 (22.3) , Loss/total: 0.48692 , Loss/sl_clf: 0.00459 , Loss/bbox: 0.03544 , Loss/giou: 0.12046 , bbox_iou: 0.88211 , grad_norm: 44.02397 , reward: 0.00112 , e_mIoU: 0.86508 , mIoU: 0.86396 , mIoU10: 0.85898 , mIoU100: 0.84035
[train: 120, 8 / 125] FPS: 16.1 (14.1) , Loss/total: 0.46262 , Loss/sl_clf: 0.00332 , Loss/bbox: 0.03514 , Loss/giou: 0.11853 , bbox_iou: 0.88388 , grad_norm: 44.58874 , reward: 0.00076 , e_mIoU: 0.86839 , mIoU: 0.86763 , mIoU10: 0.86028 , mIoU100: 0.84050
[train: 120, 9 / 125] FPS: 16.8 (24.7) , Loss/total: 0.71827 , Loss/sl_clf: 0.01991 , Loss/bbox: 0.03558 , Loss/giou: 0.12088 , bbox_iou: 0.88162 , grad_norm: 43.34234 , reward: 0.00528 , e_mIoU: 0.86630 , mIoU: 0.86101 , mIoU10: 0.86081 , mIoU100: 0.84061
[train: 120, 10 / 125] FPS: 16.9 (18.0) , Loss/total: 0.65189 , Loss/sl_clf: 0.01451 , Loss/bbox: 0.03729 , Loss/giou: 0.12393 , bbox_iou: 0.87887 , grad_norm: 43.36047 , reward: 0.00370 , e_mIoU: 0.86312 , mIoU: 0.85942 , mIoU10: 0.86067 , mIoU100: 0.84081
[train: 120, 11 / 125] FPS: 17.3 (22.0) , Loss/total: 0.63498 , Loss/sl_clf: 0.01275 , Loss/bbox: 0.03802 , Loss/giou: 0.12679 , bbox_iou: 0.87631 , grad_norm: 43.19478 , reward: 0.00325 , e_mIoU: 0.86049 , mIoU: 0.85724 , mIoU10: 0.86091 , mIoU100: 0.84103
[train: 120, 12 / 125] FPS: 17.4 (19.4) , Loss/total: 0.64611 , Loss/sl_clf: 0.01234 , Loss/bbox: 0.04089 , Loss/giou: 0.12831 , bbox_iou: 0.87465 , grad_norm: 43.26185 , reward: 0.00318 , e_mIoU: 0.85854 , mIoU: 0.85536 , mIoU10: 0.86091 , mIoU100: 0.84124
[train: 120, 13 / 125] FPS: 17.9 (26.3) , Loss/total: 0.65784 , Loss/sl_clf: 0.01181 , Loss/bbox: 0.04290 , Loss/giou: 0.13313 , bbox_iou: 0.87070 , grad_norm: 44.90731 , reward: 0.00312 , e_mIoU: 0.85462 , mIoU: 0.85151 , mIoU10: 0.86051 , mIoU100: 0.84139
[train: 120, 14 / 125] FPS: 18.4 (29.5) , Loss/total: 0.61726 , Loss/sl_clf: 0.00997 , Loss/bbox: 0.04161 , Loss/giou: 0.12981 , bbox_iou: 0.87385 , grad_norm: 43.76104 , reward: 0.00263 , e_mIoU: 0.85849 , mIoU: 0.85586 , mIoU10: 0.86044 , mIoU100: 0.84160
[train: 120, 15 / 125] FPS: 18.0 (14.1) , Loss/total: 0.58961 , Loss/sl_clf: 0.00934 , Loss/bbox: 0.03987 , Loss/giou: 0.12509 , bbox_iou: 0.87835 , grad_norm: 42.86701 , reward: 0.00247 , e_mIoU: 0.86398 , mIoU: 0.86151 , mIoU10: 0.86095 , mIoU100: 0.84185
[train: 120, 16 / 125] FPS: 18.2 (22.1) , Loss/total: 0.57445 , Loss/sl_clf: 0.00861 , Loss/bbox: 0.03950 , Loss/giou: 0.12387 , bbox_iou: 0.87941 , grad_norm: 43.42530 , reward: 0.00227 , e_mIoU: 0.86530 , mIoU: 0.86302 , mIoU10: 0.86126 , mIoU100: 0.84206
[train: 120, 17 / 125] FPS: 18.7 (32.1) , Loss/total: 0.69059 , Loss/sl_clf: 0.01650 , Loss/bbox: 0.03914 , Loss/giou: 0.12365 , bbox_iou: 0.87954 , grad_norm: 46.00386 , reward: 0.00462 , e_mIoU: 0.86621 , mIoU: 0.86159 , mIoU10: 0.86118 , mIoU100: 0.84223
[train: 120, 18 / 125] FPS: 18.9 (22.7) , Loss/total: 0.53347 , Loss/sl_clf: 0.00656 , Loss/bbox: 0.03835 , Loss/giou: 0.12164 , bbox_iou: 0.88142 , grad_norm: 48.11228 , reward: 0.00193 , e_mIoU: 0.85908 , mIoU: 0.85715 , mIoU10: 0.86049 , mIoU100: 0.84234
[train: 120, 19 / 125] FPS: 19.0 (21.2) , Loss/total: 0.51396 , Loss/sl_clf: 0.00556 , Loss/bbox: 0.03772 , Loss/giou: 0.12095 , bbox_iou: 0.88202 , grad_norm: 47.47699 , reward: 0.00165 , e_mIoU: 0.86042 , mIoU: 0.85878 , mIoU10: 0.86030 , mIoU100: 0.84246
[train: 120, 20 / 125] FPS: 19.3 (29.0) , Loss/total: 0.50075 , Loss/sl_clf: 0.00510 , Loss/bbox: 0.03715 , Loss/giou: 0.11927 , bbox_iou: 0.88358 , grad_norm: 46.59342 , reward: 0.00150 , e_mIoU: 0.86277 , mIoU: 0.86126 , mIoU10: 0.86044 , mIoU100: 0.84261
[train: 120, 21 / 125] FPS: 19.6 (27.7) , Loss/total: 0.49787 , Loss/sl_clf: 0.00466 , Loss/bbox: 0.03780 , Loss/giou: 0.11951 , bbox_iou: 0.88395 , grad_norm: 46.21938 , reward: 0.00137 , e_mIoU: 0.85907 , mIoU: 0.85769 , mIoU10: 0.86033 , mIoU100: 0.84272
[train: 120, 22 / 125] FPS: 19.9 (30.6) , Loss/total: 0.49883 , Loss/sl_clf: 0.00430 , Loss/bbox: 0.03837 , Loss/giou: 0.12128 , bbox_iou: 0.88218 , grad_norm: 45.48214 , reward: 0.00126 , e_mIoU: 0.85814 , mIoU: 0.85687 , mIoU10: 0.86026 , mIoU100: 0.84289
[train: 120, 23 / 125] FPS: 20.3 (32.9) , Loss/total: 0.50214 , Loss/sl_clf: 0.00393 , Loss/bbox: 0.03931 , Loss/giou: 0.12332 , bbox_iou: 0.88029 , grad_norm: 44.33051 , reward: 0.00115 , e_mIoU: 0.85157 , mIoU: 0.85042 , mIoU10: 0.85977 , mIoU100: 0.84300
[train: 120, 24 / 125] FPS: 20.4 (22.2) , Loss/total: 0.49680 , Loss/sl_clf: 0.00358 , Loss/bbox: 0.03920 , Loss/giou: 0.12354 , bbox_iou: 0.88008 , grad_norm: 43.84177 , reward: 0.00103 , e_mIoU: 0.85157 , mIoU: 0.85053 , mIoU10: 0.85907 , mIoU100: 0.84312
[train: 120, 25 / 125] FPS: 20.2 (17.7) , Loss/total: 0.49790 , Loss/sl_clf: 0.00347 , Loss/bbox: 0.03945 , Loss/giou: 0.12428 , bbox_iou: 0.87943 , grad_norm: 43.53260 , reward: 0.00102 , e_mIoU: 0.85108 , mIoU: 0.85007 , mIoU10: 0.85803 , mIoU100: 0.84325
[train: 120, 26 / 125] FPS: 20.1 (16.9) , Loss/total: 0.48965 , Loss/sl_clf: 0.00322 , Loss/bbox: 0.03898 , Loss/giou: 0.12321 , bbox_iou: 0.88040 , grad_norm: 43.90318 , reward: 0.00093 , e_mIoU: 0.85296 , mIoU: 0.85204 , mIoU10: 0.85712 , mIoU100: 0.84343
[train: 120, 27 / 125] FPS: 19.9 (15.9) , Loss/total: 0.46210 , Loss/sl_clf: 0.00133 , Loss/bbox: 0.03900 , Loss/giou: 0.12358 , bbox_iou: 0.88006 , grad_norm: 44.43294 , reward: 0.00038 , e_mIoU: 0.85300 , mIoU: 0.85262 , mIoU10: 0.85639 , mIoU100: 0.84359
[train: 120, 28 / 125] FPS: 20.1 (25.2) , Loss/total: 0.46060 , Loss/sl_clf: 0.00140 , Loss/bbox: 0.03868 , Loss/giou: 0.12307 , bbox_iou: 0.88050 , grad_norm: 43.81959 , reward: 0.00041 , e_mIoU: 0.85400 , mIoU: 0.85360 , mIoU10: 0.85606 , mIoU100: 0.84376
[train: 120, 29 / 125] FPS: 20.4 (36.0) , Loss/total: 0.45454 , Loss/sl_clf: 0.00122 , Loss/bbox: 0.03834 , Loss/giou: 0.12229 , bbox_iou: 0.88119 , grad_norm: 43.74290 , reward: 0.00034 , e_mIoU: 0.85538 , mIoU: 0.85503 , mIoU10: 0.85578 , mIoU100: 0.84393
[train: 120, 30 / 125] FPS: 20.6 (33.3) , Loss/total: 0.46138 , Loss/sl_clf: 0.00126 , Loss/bbox: 0.03924 , Loss/giou: 0.12310 , bbox_iou: 0.88089 , grad_norm: 44.08302 , reward: 0.00035 , e_mIoU: 0.85413 , mIoU: 0.85378 , mIoU10: 0.85521 , mIoU100: 0.84405
[train: 120, 31 / 125] FPS: 20.5 (16.9) , Loss/total: 0.46240 , Loss/sl_clf: 0.00119 , Loss/bbox: 0.03943 , Loss/giou: 0.12372 , bbox_iou: 0.88032 , grad_norm: 43.80712 , reward: 0.00035 , e_mIoU: 0.85378 , mIoU: 0.85343 , mIoU10: 0.85487 , mIoU100: 0.84418
[train: 120, 32 / 125] FPS: 20.5 (19.9) , Loss/total: 0.40131 , Loss/sl_clf: -0.00334 , Loss/bbox: 0.04009 , Loss/giou: 0.12550 , bbox_iou: 0.87857 , grad_norm: 44.34441 , reward: -0.00082 , e_mIoU: 0.85116 , mIoU: 0.85198 , mIoU10: 0.85444 , mIoU100: 0.84428
[train: 120, 33 / 125] FPS: 20.7 (35.1) , Loss/total: 0.41326 , Loss/sl_clf: -0.00327 , Loss/bbox: 0.04105 , Loss/giou: 0.12848 , bbox_iou: 0.87574 , grad_norm: 44.37906 , reward: -0.00079 , e_mIoU: 0.84584 , mIoU: 0.84663 , mIoU10: 0.85394 , mIoU100: 0.84432
[train: 120, 34 / 125] FPS: 20.8 (22.3) , Loss/total: 0.41867 , Loss/sl_clf: -0.00291 , Loss/bbox: 0.04112 , Loss/giou: 0.12835 , bbox_iou: 0.87583 , grad_norm: 44.72159 , reward: -0.00069 , e_mIoU: 0.84646 , mIoU: 0.84715 , mIoU10: 0.85350 , mIoU100: 0.84434
[train: 120, 35 / 125] FPS: 20.8 (21.5) , Loss/total: 0.41086 , Loss/sl_clf: -0.00365 , Loss/bbox: 0.04134 , Loss/giou: 0.12945 , bbox_iou: 0.87493 , grad_norm: 44.52935 , reward: -0.00089 , e_mIoU: 0.84394 , mIoU: 0.84484 , mIoU10: 0.85288 , mIoU100: 0.84436
[train: 120, 36 / 125] FPS: 21.0 (31.4) , Loss/total: 0.32182 , Loss/sl_clf: -0.01037 , Loss/bbox: 0.04272 , Loss/giou: 0.13191 , bbox_iou: 0.87316 , grad_norm: 46.05547 , reward: -0.00295 , e_mIoU: 0.84138 , mIoU: 0.84433 , mIoU10: 0.85209 , mIoU100: 0.84436
[train: 120, 37 / 125] FPS: 21.0 (20.6) , Loss/total: 0.32469 , Loss/sl_clf: -0.01017 , Loss/bbox: 0.04264 , Loss/giou: 0.13199 , bbox_iou: 0.87300 , grad_norm: 45.74641 , reward: -0.00288 , e_mIoU: 0.84090 , mIoU: 0.84378 , mIoU10: 0.85122 , mIoU100: 0.84434
[train: 120, 38 / 125] FPS: 21.1 (30.6) , Loss/total: 0.33140 , Loss/sl_clf: -0.00986 , Loss/bbox: 0.04284 , Loss/giou: 0.13254 , bbox_iou: 0.87247 , grad_norm: 45.58754 , reward: -0.00279 , e_mIoU: 0.84108 , mIoU: 0.84387 , mIoU10: 0.85031 , mIoU100: 0.84434
[train: 120, 39 / 125] FPS: 21.3 (28.2) , Loss/total: 0.34287 , Loss/sl_clf: -0.00952 , Loss/bbox: 0.04360 , Loss/giou: 0.13388 , bbox_iou: 0.87120 , grad_norm: 45.77084 , reward: -0.00270 , e_mIoU: 0.84065 , mIoU: 0.84335 , mIoU10: 0.84926 , mIoU100: 0.84435
[train: 120, 40 / 125] FPS: 21.2 (17.2) , Loss/total: 0.34565 , Loss/sl_clf: -0.00955 , Loss/bbox: 0.04375 , Loss/giou: 0.13505 , bbox_iou: 0.87001 , grad_norm: 45.80618 , reward: -0.00271 , e_mIoU: 0.83943 , mIoU: 0.84214 , mIoU10: 0.84821 , mIoU100: 0.84435
[train: 120, 41 / 125] FPS: 21.3 (30.3) , Loss/total: 0.34129 , Loss/sl_clf: -0.00965 , Loss/bbox: 0.04348 , Loss/giou: 0.13431 , bbox_iou: 0.87066 , grad_norm: 45.47709 , reward: -0.00272 , e_mIoU: 0.84033 , mIoU: 0.84306 , mIoU10: 0.84730 , mIoU100: 0.84435
[train: 120, 42 / 125] FPS: 21.2 (17.7) , Loss/total: 0.28334 , Loss/sl_clf: -0.01372 , Loss/bbox: 0.04419 , Loss/giou: 0.13413 , bbox_iou: 0.87138 , grad_norm: 46.66088 , reward: -0.00395 , e_mIoU: 0.83819 , mIoU: 0.84215 , mIoU10: 0.84643 , mIoU100: 0.84434
[train: 120, 43 / 125] FPS: 21.2 (20.8) , Loss/total: 0.28741 , Loss/sl_clf: -0.01360 , Loss/bbox: 0.04442 , Loss/giou: 0.13461 , bbox_iou: 0.87096 , grad_norm: 46.38076 , reward: -0.00392 , e_mIoU: 0.83803 , mIoU: 0.84195 , mIoU10: 0.84596 , mIoU100: 0.84435
[train: 120, 44 / 125] FPS: 21.3 (26.0) , Loss/total: 0.28960 , Loss/sl_clf: -0.01348 , Loss/bbox: 0.04441 , Loss/giou: 0.13483 , bbox_iou: 0.87070 , grad_norm: 46.33379 , reward: -0.00388 , e_mIoU: 0.83815 , mIoU: 0.84203 , mIoU10: 0.84548 , mIoU100: 0.84436
[train: 120, 45 / 125] FPS: 21.2 (16.9) , Loss/total: 0.28706 , Loss/sl_clf: -0.01355 , Loss/bbox: 0.04424 , Loss/giou: 0.13452 , bbox_iou: 0.87096 , grad_norm: 46.43205 , reward: -0.00391 , e_mIoU: 0.83891 , mIoU: 0.84281 , mIoU10: 0.84526 , mIoU100: 0.84436
[train: 120, 46 / 125] FPS: 21.3 (34.7) , Loss/total: 0.28570 , Loss/sl_clf: -0.01331 , Loss/bbox: 0.04376 , Loss/giou: 0.13326 , bbox_iou: 0.87211 , grad_norm: 46.03279 , reward: -0.00384 , e_mIoU: 0.84071 , mIoU: 0.84455 , mIoU10: 0.84526 , mIoU100: 0.84437
[train: 120, 47 / 125] FPS: 21.4 (22.9) , Loss/total: 0.28632 , Loss/sl_clf: -0.01381 , Loss/bbox: 0.04470 , Loss/giou: 0.13497 , bbox_iou: 0.87060 , grad_norm: 46.14822 , reward: -0.00401 , e_mIoU: 0.83702 , mIoU: 0.84103 , mIoU10: 0.84496 , mIoU100: 0.84434
[train: 120, 48 / 125] FPS: 20.9 (10.1) , Loss/total: 0.28858 , Loss/sl_clf: -0.01344 , Loss/bbox: 0.04438 , Loss/giou: 0.13413 , bbox_iou: 0.87140 , grad_norm: 45.99886 , reward: -0.00391 , e_mIoU: 0.83857 , mIoU: 0.84248 , mIoU10: 0.84479 , mIoU100: 0.84433
[train: 120, 49 / 125] FPS: 21.1 (32.7) , Loss/total: 0.28820 , Loss/sl_clf: -0.01334 , Loss/bbox: 0.04419 , Loss/giou: 0.13367 , bbox_iou: 0.87179 , grad_norm: 46.03891 , reward: -0.00388 , e_mIoU: 0.83937 , mIoU: 0.84325 , mIoU10: 0.84476 , mIoU100: 0.84433
[train: 120, 50 / 125] FPS: 21.1 (21.1) , Loss/total: 0.29389 , Loss/sl_clf: -0.01300 , Loss/bbox: 0.04424 , Loss/giou: 0.13387 , bbox_iou: 0.87157 , grad_norm: 46.01740 , reward: -0.00378 , e_mIoU: 0.83937 , mIoU: 0.84315 , mIoU10: 0.84481 , mIoU100: 0.84433
[train: 120, 51 / 125] FPS: 21.1 (25.8) , Loss/total: 0.28337 , Loss/sl_clf: -0.01370 , Loss/bbox: 0.04415 , Loss/giou: 0.13408 , bbox_iou: 0.87132 , grad_norm: 45.91993 , reward: -0.00398 , e_mIoU: 0.83701 , mIoU: 0.84098 , mIoU10: 0.84456 , mIoU100: 0.84429
[train: 120, 52 / 125] FPS: 21.2 (26.0) , Loss/total: 0.27251 , Loss/sl_clf: -0.01455 , Loss/bbox: 0.04435 , Loss/giou: 0.13448 , bbox_iou: 0.87099 , grad_norm: 45.86866 , reward: -0.00420 , e_mIoU: 0.83645 , mIoU: 0.84065 , mIoU10: 0.84437 , mIoU100: 0.84425
[train: 120, 53 / 125] FPS: 21.3 (27.7) , Loss/total: 0.27497 , Loss/sl_clf: -0.01427 , Loss/bbox: 0.04415 , Loss/giou: 0.13416 , bbox_iou: 0.87125 , grad_norm: 45.41065 , reward: -0.00412 , e_mIoU: 0.83727 , mIoU: 0.84140 , mIoU10: 0.84427 , mIoU100: 0.84422
[train: 120, 54 / 125] FPS: 21.4 (25.8) , Loss/total: 0.27603 , Loss/sl_clf: -0.01403 , Loss/bbox: 0.04388 , Loss/giou: 0.13355 , bbox_iou: 0.87177 , grad_norm: 45.41851 , reward: -0.00405 , e_mIoU: 0.83836 , mIoU: 0.84241 , mIoU10: 0.84426 , mIoU100: 0.84419
[train: 120, 55 / 125] FPS: 21.3 (20.1) , Loss/total: 0.27880 , Loss/sl_clf: -0.01380 , Loss/bbox: 0.04379 , Loss/giou: 0.13341 , bbox_iou: 0.87188 , grad_norm: 45.77953 , reward: -0.00398 , e_mIoU: 0.83900 , mIoU: 0.84298 , mIoU10: 0.84425 , mIoU100: 0.84417
[train: 120, 56 / 125] FPS: 21.3 (19.4) , Loss/total: 0.27868 , Loss/sl_clf: -0.01362 , Loss/bbox: 0.04346 , Loss/giou: 0.13282 , bbox_iou: 0.87240 , grad_norm: 45.63506 , reward: -0.00394 , e_mIoU: 0.83993 , mIoU: 0.84387 , mIoU10: 0.84419 , mIoU100: 0.84414
[train: 120, 57 / 125] FPS: 21.4 (26.5) , Loss/total: 0.27949 , Loss/sl_clf: -0.01339 , Loss/bbox: 0.04320 , Loss/giou: 0.13214 , bbox_iou: 0.87299 , grad_norm: 45.75445 , reward: -0.00388 , e_mIoU: 0.84109 , mIoU: 0.84497 , mIoU10: 0.84453 , mIoU100: 0.84412
[train: 120, 58 / 125] FPS: 21.4 (20.8) , Loss/total: 0.28339 , Loss/sl_clf: -0.01306 , Loss/bbox: 0.04312 , Loss/giou: 0.13183 , bbox_iou: 0.87326 , grad_norm: 45.91810 , reward: -0.00378 , e_mIoU: 0.84189 , mIoU: 0.84567 , mIoU10: 0.84481 , mIoU100: 0.84409
[train: 120, 59 / 125] FPS: 21.2 (14.2) , Loss/total: 0.28392 , Loss/sl_clf: -0.01287 , Loss/bbox: 0.04288 , Loss/giou: 0.13125 , bbox_iou: 0.87377 , grad_norm: 46.05063 , reward: -0.00374 , e_mIoU: 0.84286 , mIoU: 0.84660 , mIoU10: 0.84512 , mIoU100: 0.84409
[train: 120, 60 / 125] FPS: 21.2 (22.0) , Loss/total: 0.36515 , Loss/sl_clf: -0.00731 , Loss/bbox: 0.04266 , Loss/giou: 0.13071 , bbox_iou: 0.87428 , grad_norm: 48.13029 , reward: -0.00224 , e_mIoU: 0.84360 , mIoU: 0.84584 , mIoU10: 0.84536 , mIoU100: 0.84408
[train: 120, 61 / 125] FPS: 21.3 (26.5) , Loss/total: 0.36724 , Loss/sl_clf: -0.00713 , Loss/bbox: 0.04260 , Loss/giou: 0.13063 , bbox_iou: 0.87434 , grad_norm: 48.04942 , reward: -0.00219 , e_mIoU: 0.84328 , mIoU: 0.84547 , mIoU10: 0.84573 , mIoU100: 0.84409
[train: 120, 62 / 125] FPS: 21.4 (30.8) , Loss/total: 0.35568 , Loss/sl_clf: -0.00797 , Loss/bbox: 0.04273 , Loss/giou: 0.13079 , bbox_iou: 0.87414 , grad_norm: 47.71792 , reward: -0.00243 , e_mIoU: 0.84332 , mIoU: 0.84575 , mIoU10: 0.84616 , mIoU100: 0.84410
[train: 120, 63 / 125] FPS: 21.5 (32.9) , Loss/total: 0.36364 , Loss/sl_clf: -0.00770 , Loss/bbox: 0.04308 , Loss/giou: 0.13184 , bbox_iou: 0.87313 , grad_norm: 47.68160 , reward: -0.00235 , e_mIoU: 0.84255 , mIoU: 0.84490 , mIoU10: 0.84644 , mIoU100: 0.84412
[train: 120, 64 / 125] FPS: 21.5 (21.2) , Loss/total: 0.36932 , Loss/sl_clf: -0.00756 , Loss/bbox: 0.04348 , Loss/giou: 0.13263 , bbox_iou: 0.87233 , grad_norm: 47.43177 , reward: -0.00231 , e_mIoU: 0.84229 , mIoU: 0.84460 , mIoU10: 0.84659 , mIoU100: 0.84415
[train: 120, 65 / 125] FPS: 21.6 (32.3) , Loss/total: 0.37605 , Loss/sl_clf: -0.00748 , Loss/bbox: 0.04421 , Loss/giou: 0.13364 , bbox_iou: 0.87165 , grad_norm: 47.20571 , reward: -0.00228 , e_mIoU: 0.84109 , mIoU: 0.84337 , mIoU10: 0.84657 , mIoU100: 0.84416
[train: 120, 66 / 125] FPS: 21.5 (18.2) , Loss/total: 0.37459 , Loss/sl_clf: -0.00735 , Loss/bbox: 0.04385 , Loss/giou: 0.13283 , bbox_iou: 0.87240 , grad_norm: 47.05834 , reward: -0.00224 , e_mIoU: 0.84229 , mIoU: 0.84453 , mIoU10: 0.84660 , mIoU100: 0.84417
[train: 120, 67 / 125] FPS: 21.6 (25.2) , Loss/total: 0.37074 , Loss/sl_clf: -0.00812 , Loss/bbox: 0.04473 , Loss/giou: 0.13441 , bbox_iou: 0.87092 , grad_norm: 47.36792 , reward: -0.00244 , e_mIoU: 0.84062 , mIoU: 0.84306 , mIoU10: 0.84638 , mIoU100: 0.84416
[train: 120, 68 / 125] FPS: 21.5 (16.9) , Loss/total: 0.37645 , Loss/sl_clf: -0.00789 , Loss/bbox: 0.04494 , Loss/giou: 0.13509 , bbox_iou: 0.87022 , grad_norm: 47.04243 , reward: -0.00238 , e_mIoU: 0.84039 , mIoU: 0.84277 , mIoU10: 0.84608 , mIoU100: 0.84414
[train: 120, 69 / 125] FPS: 21.3 (14.1) , Loss/total: 0.37910 , Loss/sl_clf: -0.00761 , Loss/bbox: 0.04477 , Loss/giou: 0.13468 , bbox_iou: 0.87059 , grad_norm: 46.90489 , reward: -0.00229 , e_mIoU: 0.84045 , mIoU: 0.84274 , mIoU10: 0.84571 , mIoU100: 0.84411
[train: 120, 70 / 125] FPS: 21.4 (31.7) , Loss/total: 0.37724 , Loss/sl_clf: -0.00756 , Loss/bbox: 0.04449 , Loss/giou: 0.13411 , bbox_iou: 0.87111 , grad_norm: 46.71570 , reward: -0.00228 , e_mIoU: 0.84127 , mIoU: 0.84355 , mIoU10: 0.84548 , mIoU100: 0.84408
[train: 120, 71 / 125] FPS: 21.4 (16.9) , Loss/total: 0.39174 , Loss/sl_clf: -0.00658 , Loss/bbox: 0.04449 , Loss/giou: 0.13397 , bbox_iou: 0.87120 , grad_norm: 46.97401 , reward: -0.00198 , e_mIoU: 0.84159 , mIoU: 0.84357 , mIoU10: 0.84529 , mIoU100: 0.84406
[train: 120, 72 / 125] FPS: 21.3 (18.4) , Loss/total: 0.39016 , Loss/sl_clf: -0.00659 , Loss/bbox: 0.04433 , Loss/giou: 0.13365 , bbox_iou: 0.87148 , grad_norm: 46.90411 , reward: -0.00198 , e_mIoU: 0.84209 , mIoU: 0.84407 , mIoU10: 0.84513 , mIoU100: 0.84403
[train: 120, 73 / 125] FPS: 21.4 (34.0) , Loss/total: 0.39060 , Loss/sl_clf: -0.00659 , Loss/bbox: 0.04435 , Loss/giou: 0.13385 , bbox_iou: 0.87128 , grad_norm: 46.74831 , reward: -0.00198 , e_mIoU: 0.84207 , mIoU: 0.84404 , mIoU10: 0.84504 , mIoU100: 0.84400
[train: 120, 74 / 125] FPS: 21.2 (12.8) , Loss/total: 0.39218 , Loss/sl_clf: -0.00635 , Loss/bbox: 0.04410 , Loss/giou: 0.13343 , bbox_iou: 0.87167 , grad_norm: 46.77439 , reward: -0.00191 , e_mIoU: 0.84209 , mIoU: 0.84400 , mIoU10: 0.84497 , mIoU100: 0.84397
[train: 120, 75 / 125] FPS: 21.3 (26.3) , Loss/total: 0.39160 , Loss/sl_clf: -0.00643 , Loss/bbox: 0.04415 , Loss/giou: 0.13368 , bbox_iou: 0.87138 , grad_norm: 46.84555 , reward: -0.00194 , e_mIoU: 0.84208 , mIoU: 0.84402 , mIoU10: 0.84501 , mIoU100: 0.84394
[train: 120, 76 / 125] FPS: 21.3 (26.9) , Loss/total: 0.44312 , Loss/sl_clf: -0.00298 , Loss/bbox: 0.04413 , Loss/giou: 0.13356 , bbox_iou: 0.87152 , grad_norm: 46.83838 , reward: -0.00101 , e_mIoU: 0.84222 , mIoU: 0.84323 , mIoU10: 0.84488 , mIoU100: 0.84391
[train: 120, 77 / 125] FPS: 21.4 (28.9) , Loss/total: 0.43257 , Loss/sl_clf: -0.00367 , Loss/bbox: 0.04406 , Loss/giou: 0.13366 , bbox_iou: 0.87139 , grad_norm: 46.99710 , reward: -0.00114 , e_mIoU: 0.84112 , mIoU: 0.84226 , mIoU10: 0.84477 , mIoU100: 0.84388
[train: 120, 78 / 125] FPS: 21.5 (27.8) , Loss/total: 0.43271 , Loss/sl_clf: -0.00364 , Loss/bbox: 0.04401 , Loss/giou: 0.13362 , bbox_iou: 0.87139 , grad_norm: 47.02110 , reward: -0.00113 , e_mIoU: 0.84142 , mIoU: 0.84255 , mIoU10: 0.84473 , mIoU100: 0.84386
[train: 120, 79 / 125] FPS: 21.4 (18.2) , Loss/total: 0.44497 , Loss/sl_clf: -0.00353 , Loss/bbox: 0.04567 , Loss/giou: 0.13478 , bbox_iou: 0.87089 , grad_norm: 46.86891 , reward: -0.00110 , e_mIoU: 0.84107 , mIoU: 0.84217 , mIoU10: 0.84464 , mIoU100: 0.84386
[train: 120, 80 / 125] FPS: 21.4 (19.9) , Loss/total: 0.44278 , Loss/sl_clf: -0.00372 , Loss/bbox: 0.04565 , Loss/giou: 0.13515 , bbox_iou: 0.87049 , grad_norm: 46.88723 , reward: -0.00115 , e_mIoU: 0.84096 , mIoU: 0.84212 , mIoU10: 0.84449 , mIoU100: 0.84385
[train: 120, 81 / 125] FPS: 21.3 (16.5) , Loss/total: 0.44768 , Loss/sl_clf: -0.00354 , Loss/bbox: 0.04577 , Loss/giou: 0.13594 , bbox_iou: 0.86971 , grad_norm: 47.18306 , reward: -0.00109 , e_mIoU: 0.84021 , mIoU: 0.84130 , mIoU10: 0.84425 , mIoU100: 0.84383
[train: 120, 82 / 125] FPS: 21.4 (24.5) , Loss/total: 0.46211 , Loss/sl_clf: -0.00272 , Loss/bbox: 0.04600 , Loss/giou: 0.13645 , bbox_iou: 0.86925 , grad_norm: 47.06617 , reward: -0.00088 , e_mIoU: 0.83940 , mIoU: 0.84028 , mIoU10: 0.84387 , mIoU100: 0.84380
[train: 120, 83 / 125] FPS: 21.4 (24.5) , Loss/total: 0.48642 , Loss/sl_clf: -0.00148 , Loss/bbox: 0.04657 , Loss/giou: 0.13788 , bbox_iou: 0.86797 , grad_norm: 47.41314 , reward: -0.00057 , e_mIoU: 0.83791 , mIoU: 0.83847 , mIoU10: 0.84331 , mIoU100: 0.84376
[train: 120, 84 / 125] FPS: 21.5 (33.4) , Loss/total: 0.46295 , Loss/sl_clf: -0.00290 , Loss/bbox: 0.04635 , Loss/giou: 0.13733 , bbox_iou: 0.86845 , grad_norm: 47.71091 , reward: -0.00096 , e_mIoU: 0.83815 , mIoU: 0.83911 , mIoU10: 0.84283 , mIoU100: 0.84373
[train: 120, 85 / 125] FPS: 21.5 (24.3) , Loss/total: 0.46296 , Loss/sl_clf: -0.00287 , Loss/bbox: 0.04627 , Loss/giou: 0.13730 , bbox_iou: 0.86849 , grad_norm: 47.70833 , reward: -0.00095 , e_mIoU: 0.83850 , mIoU: 0.83945 , mIoU10: 0.84239 , mIoU100: 0.84371
[train: 120, 86 / 125] FPS: 21.5 (18.2) , Loss/total: 0.46678 , Loss/sl_clf: -0.00264 , Loss/bbox: 0.04627 , Loss/giou: 0.13749 , bbox_iou: 0.86828 , grad_norm: 47.63589 , reward: -0.00087 , e_mIoU: 0.83805 , mIoU: 0.83893 , mIoU10: 0.84197 , mIoU100: 0.84368
[train: 120, 87 / 125] FPS: 21.5 (20.6) , Loss/total: 0.43891 , Loss/sl_clf: -0.00440 , Loss/bbox: 0.04614 , Loss/giou: 0.13712 , bbox_iou: 0.86860 , grad_norm: 47.68760 , reward: -0.00133 , e_mIoU: 0.83788 , mIoU: 0.83921 , mIoU10: 0.84167 , mIoU100: 0.84365
[train: 120, 88 / 125] FPS: 21.4 (19.8) , Loss/total: 0.43870 , Loss/sl_clf: -0.00445 , Loss/bbox: 0.04614 , Loss/giou: 0.13733 , bbox_iou: 0.86840 , grad_norm: 47.59709 , reward: -0.00135 , e_mIoU: 0.83727 , mIoU: 0.83862 , mIoU10: 0.84129 , mIoU100: 0.84363
[train: 120, 89 / 125] FPS: 21.5 (35.5) , Loss/total: 0.44797 , Loss/sl_clf: -0.00381 , Loss/bbox: 0.04607 , Loss/giou: 0.13736 , bbox_iou: 0.86834 , grad_norm: 47.64104 , reward: -0.00121 , e_mIoU: 0.83706 , mIoU: 0.83827 , mIoU10: 0.84091 , mIoU100: 0.84359
[train: 120, 90 / 125] FPS: 21.5 (19.0) , Loss/total: 0.43902 , Loss/sl_clf: -0.00430 , Loss/bbox: 0.04590 , Loss/giou: 0.13697 , bbox_iou: 0.86868 , grad_norm: 47.58022 , reward: -0.00135 , e_mIoU: 0.83759 , mIoU: 0.83894 , mIoU10: 0.84060 , mIoU100: 0.84356
[train: 120, 91 / 125] FPS: 21.4 (16.5) , Loss/total: 0.43767 , Loss/sl_clf: -0.00431 , Loss/bbox: 0.04579 , Loss/giou: 0.13669 , bbox_iou: 0.86892 , grad_norm: 47.71475 , reward: -0.00136 , e_mIoU: 0.83702 , mIoU: 0.83838 , mIoU10: 0.84032 , mIoU100: 0.84352
[train: 120, 92 / 125] FPS: 21.4 (22.3) , Loss/total: 0.43647 , Loss/sl_clf: -0.00430 , Loss/bbox: 0.04562 , Loss/giou: 0.13642 , bbox_iou: 0.86915 , grad_norm: 47.61230 , reward: -0.00135 , e_mIoU: 0.83743 , mIoU: 0.83878 , mIoU10: 0.84017 , mIoU100: 0.84349
[train: 120, 93 / 125] FPS: 21.5 (30.4) , Loss/total: 0.43514 , Loss/sl_clf: -0.00434 , Loss/bbox: 0.04550 , Loss/giou: 0.13635 , bbox_iou: 0.86919 , grad_norm: 47.56500 , reward: -0.00136 , e_mIoU: 0.83777 , mIoU: 0.83914 , mIoU10: 0.84021 , mIoU100: 0.84347
[train: 120, 94 / 125] FPS: 21.5 (18.0) , Loss/total: 0.43785 , Loss/sl_clf: -0.00417 , Loss/bbox: 0.04551 , Loss/giou: 0.13644 , bbox_iou: 0.86907 , grad_norm: 47.70093 , reward: -0.00131 , e_mIoU: 0.83783 , mIoU: 0.83914 , mIoU10: 0.84021 , mIoU100: 0.84343
[train: 120, 95 / 125] FPS: 21.5 (21.4) , Loss/total: 0.43588 , Loss/sl_clf: -0.00414 , Loss/bbox: 0.04526 , Loss/giou: 0.13586 , bbox_iou: 0.86960 , grad_norm: 47.75829 , reward: -0.00130 , e_mIoU: 0.83864 , mIoU: 0.83994 , mIoU10: 0.84025 , mIoU100: 0.84342
[train: 120, 96 / 125] FPS: 21.3 (14.1) , Loss/total: 0.43646 , Loss/sl_clf: -0.00408 , Loss/bbox: 0.04527 , Loss/giou: 0.13568 , bbox_iou: 0.86974 , grad_norm: 47.78666 , reward: -0.00128 , e_mIoU: 0.83915 , mIoU: 0.84043 , mIoU10: 0.84038 , mIoU100: 0.84340
[train: 120, 97 / 125] FPS: 21.4 (37.1) , Loss/total: 0.43819 , Loss/sl_clf: -0.00409 , Loss/bbox: 0.04538 , Loss/giou: 0.13635 , bbox_iou: 0.86907 , grad_norm: 47.77202 , reward: -0.00128 , e_mIoU: 0.83878 , mIoU: 0.84007 , mIoU10: 0.84046 , mIoU100: 0.84337
[train: 120, 98 / 125] FPS: 21.5 (32.8) , Loss/total: 0.43876 , Loss/sl_clf: -0.00418 , Loss/bbox: 0.04558 , Loss/giou: 0.13677 , bbox_iou: 0.86877 , grad_norm: 47.72337 , reward: -0.00129 , e_mIoU: 0.83778 , mIoU: 0.83907 , mIoU10: 0.84048 , mIoU100: 0.84334
[train: 120, 99 / 125] FPS: 21.4 (12.8) , Loss/total: 0.42776 , Loss/sl_clf: -0.00481 , Loss/bbox: 0.04541 , Loss/giou: 0.13640 , bbox_iou: 0.86911 , grad_norm: 47.93720 , reward: -0.00144 , e_mIoU: 0.83804 , mIoU: 0.83949 , mIoU10: 0.84058 , mIoU100: 0.84330
[train: 120, 100 / 125] FPS: 21.5 (33.7) , Loss/total: 0.42876 , Loss/sl_clf: -0.00470 , Loss/bbox: 0.04535 , Loss/giou: 0.13630 , bbox_iou: 0.86918 , grad_norm: 47.75518 , reward: -0.00141 , e_mIoU: 0.83803 , mIoU: 0.83944 , mIoU10: 0.84062 , mIoU100: 0.84326
[train: 120, 101 / 125] FPS: 21.4 (17.7) , Loss/total: 0.42682 , Loss/sl_clf: -0.00476 , Loss/bbox: 0.04522 , Loss/giou: 0.13609 , bbox_iou: 0.86935 , grad_norm: 47.75588 , reward: -0.00143 , e_mIoU: 0.83845 , mIoU: 0.83988 , mIoU10: 0.84074 , mIoU100: 0.84324
[train: 120, 102 / 125] FPS: 21.4 (22.7) , Loss/total: 0.42676 , Loss/sl_clf: -0.00485 , Loss/bbox: 0.04532 , Loss/giou: 0.13642 , bbox_iou: 0.86905 , grad_norm: 47.76818 , reward: -0.00146 , e_mIoU: 0.83811 , mIoU: 0.83956 , mIoU10: 0.84080 , mIoU100: 0.84320
[train: 120, 103 / 125] FPS: 21.5 (32.1) , Loss/total: 0.42885 , Loss/sl_clf: -0.00475 , Loss/bbox: 0.04537 , Loss/giou: 0.13662 , bbox_iou: 0.86888 , grad_norm: 47.75878 , reward: -0.00142 , e_mIoU: 0.83800 , mIoU: 0.83942 , mIoU10: 0.84082 , mIoU100: 0.84317
[train: 120, 104 / 125] FPS: 21.5 (19.1) , Loss/total: 0.42453 , Loss/sl_clf: -0.00495 , Loss/bbox: 0.04522 , Loss/giou: 0.13634 , bbox_iou: 0.86912 , grad_norm: 47.69368 , reward: -0.00148 , e_mIoU: 0.83844 , mIoU: 0.83992 , mIoU10: 0.84088 , mIoU100: 0.84313
[train: 120, 105 / 125] FPS: 21.5 (23.6) , Loss/total: 0.40910 , Loss/sl_clf: -0.00615 , Loss/bbox: 0.04559 , Loss/giou: 0.13673 , bbox_iou: 0.86894 , grad_norm: 47.86443 , reward: -0.00186 , e_mIoU: 0.83786 , mIoU: 0.83972 , mIoU10: 0.84085 , mIoU100: 0.84310
[train: 120, 106 / 125] FPS: 21.5 (27.5) , Loss/total: 0.41126 , Loss/sl_clf: -0.00609 , Loss/bbox: 0.04571 , Loss/giou: 0.13708 , bbox_iou: 0.86858 , grad_norm: 47.78022 , reward: -0.00184 , e_mIoU: 0.83761 , mIoU: 0.83945 , mIoU10: 0.84074 , mIoU100: 0.84305
[train: 120, 107 / 125] FPS: 21.5 (20.8) , Loss/total: 0.42040 , Loss/sl_clf: -0.00546 , Loss/bbox: 0.04566 , Loss/giou: 0.13699 , bbox_iou: 0.86865 , grad_norm: 47.84992 , reward: -0.00165 , e_mIoU: 0.83788 , mIoU: 0.83953 , mIoU10: 0.84068 , mIoU100: 0.84300
[train: 120, 108 / 125] FPS: 21.5 (16.5) , Loss/total: 0.41866 , Loss/sl_clf: -0.00547 , Loss/bbox: 0.04549 , Loss/giou: 0.13659 , bbox_iou: 0.86900 , grad_norm: 47.77891 , reward: -0.00165 , e_mIoU: 0.83839 , mIoU: 0.84004 , mIoU10: 0.84077 , mIoU100: 0.84295
[train: 120, 109 / 125] FPS: 21.5 (23.1) , Loss/total: 0.42467 , Loss/sl_clf: -0.00511 , Loss/bbox: 0.04553 , Loss/giou: 0.13680 , bbox_iou: 0.86880 , grad_norm: 47.62931 , reward: -0.00155 , e_mIoU: 0.83822 , mIoU: 0.83977 , mIoU10: 0.84078 , mIoU100: 0.84291
[train: 120, 110 / 125] FPS: 21.5 (29.5) , Loss/total: 0.42066 , Loss/sl_clf: -0.00535 , Loss/bbox: 0.04548 , Loss/giou: 0.13678 , bbox_iou: 0.86881 , grad_norm: 47.67258 , reward: -0.00162 , e_mIoU: 0.83838 , mIoU: 0.84000 , mIoU10: 0.84083 , mIoU100: 0.84286
[train: 120, 111 / 125] FPS: 21.5 (23.7) , Loss/total: 0.35835 , Loss/sl_clf: -0.00953 , Loss/bbox: 0.04549 , Loss/giou: 0.13689 , bbox_iou: 0.86869 , grad_norm: 49.40823 , reward: -0.00257 , e_mIoU: 0.83768 , mIoU: 0.84025 , mIoU10: 0.84086 , mIoU100: 0.84282
[train: 120, 112 / 125] FPS: 21.5 (15.7) , Loss/total: 0.35665 , Loss/sl_clf: -0.00957 , Loss/bbox: 0.04538 , Loss/giou: 0.13660 , bbox_iou: 0.86893 , grad_norm: 49.68421 , reward: -0.00258 , e_mIoU: 0.83809 , mIoU: 0.84068 , mIoU10: 0.84096 , mIoU100: 0.84279
[train: 120, 113 / 125] FPS: 21.5 (30.6) , Loss/total: 0.35665 , Loss/sl_clf: -0.00947 , Loss/bbox: 0.04523 , Loss/giou: 0.13628 , bbox_iou: 0.86921 , grad_norm: 49.86459 , reward: -0.00256 , e_mIoU: 0.83861 , mIoU: 0.84117 , mIoU10: 0.84112 , mIoU100: 0.84276
[train: 120, 114 / 125] FPS: 21.5 (17.7) , Loss/total: 0.35756 , Loss/sl_clf: -0.00937 , Loss/bbox: 0.04517 , Loss/giou: 0.13616 , bbox_iou: 0.86929 , grad_norm: 49.72911 , reward: -0.00253 , e_mIoU: 0.83892 , mIoU: 0.84145 , mIoU10: 0.84126 , mIoU100: 0.84273
[train: 120, 115 / 125] FPS: 21.5 (20.5) , Loss/total: 0.35718 , Loss/sl_clf: -0.00928 , Loss/bbox: 0.04500 , Loss/giou: 0.13570 , bbox_iou: 0.86972 , grad_norm: 49.61376 , reward: -0.00250 , e_mIoU: 0.83956 , mIoU: 0.84206 , mIoU10: 0.84148 , mIoU100: 0.84270
[train: 120, 116 / 125] FPS: 21.5 (27.9) , Loss/total: 0.35763 , Loss/sl_clf: -0.00925 , Loss/bbox: 0.04498 , Loss/giou: 0.13572 , bbox_iou: 0.86967 , grad_norm: 49.63556 , reward: -0.00250 , e_mIoU: 0.83968 , mIoU: 0.84218 , mIoU10: 0.84174 , mIoU100: 0.84267
[train: 120, 117 / 125] FPS: 21.6 (25.8) , Loss/total: 0.35714 , Loss/sl_clf: -0.00923 , Loss/bbox: 0.04489 , Loss/giou: 0.13557 , bbox_iou: 0.86980 , grad_norm: 49.41771 , reward: -0.00250 , e_mIoU: 0.83988 , mIoU: 0.84238 , mIoU10: 0.84200 , mIoU100: 0.84264
[train: 120, 118 / 125] FPS: 21.6 (27.5) , Loss/total: 0.35620 , Loss/sl_clf: -0.00915 , Loss/bbox: 0.04467 , Loss/giou: 0.13501 , bbox_iou: 0.87032 , grad_norm: 49.13121 , reward: -0.00247 , e_mIoU: 0.84064 , mIoU: 0.84312 , mIoU10: 0.84229 , mIoU100: 0.84262
[train: 120, 119 / 125] FPS: 21.6 (21.6) , Loss/total: 0.35529 , Loss/sl_clf: -0.00908 , Loss/bbox: 0.04448 , Loss/giou: 0.13453 , bbox_iou: 0.87077 , grad_norm: 49.13800 , reward: -0.00246 , e_mIoU: 0.84127 , mIoU: 0.84372 , mIoU10: 0.84267 , mIoU100: 0.84261
[train: 120, 120 / 125] FPS: 21.6 (18.0) , Loss/total: 0.35230 , Loss/sl_clf: -0.00924 , Loss/bbox: 0.04443 , Loss/giou: 0.13439 , bbox_iou: 0.87090 , grad_norm: 49.06143 , reward: -0.00250 , e_mIoU: 0.84068 , mIoU: 0.84318 , mIoU10: 0.84296 , mIoU100: 0.84258
[train: 120, 121 / 125] FPS: 21.6 (26.4) , Loss/total: 0.34899 , Loss/sl_clf: -0.00936 , Loss/bbox: 0.04426 , Loss/giou: 0.13403 , bbox_iou: 0.87122 , grad_norm: 48.91782 , reward: -0.00254 , e_mIoU: 0.84123 , mIoU: 0.84377 , mIoU10: 0.84329 , mIoU100: 0.84257
[train: 120, 122 / 125] FPS: 21.6 (18.8) , Loss/total: 0.34794 , Loss/sl_clf: -0.00931 , Loss/bbox: 0.04407 , Loss/giou: 0.13358 , bbox_iou: 0.87164 , grad_norm: 48.91745 , reward: -0.00253 , e_mIoU: 0.84187 , mIoU: 0.84440 , mIoU10: 0.84364 , mIoU100: 0.84256
[train: 120, 123 / 125] FPS: 21.6 (22.0) , Loss/total: 0.34621 , Loss/sl_clf: -0.00935 , Loss/bbox: 0.04397 , Loss/giou: 0.13332 , bbox_iou: 0.87187 , grad_norm: 48.77783 , reward: -0.00254 , e_mIoU: 0.84225 , mIoU: 0.84479 , mIoU10: 0.84398 , mIoU100: 0.84257
[train: 120, 124 / 125] FPS: 21.6 (29.1) , Loss/total: 0.30330 , Loss/sl_clf: -0.01242 , Loss/bbox: 0.04434 , Loss/giou: 0.13398 , bbox_iou: 0.87141 , grad_norm: 49.58205 , reward: -0.00331 , e_mIoU: 0.84088 , mIoU: 0.84419 , mIoU10: 0.84424 , mIoU100: 0.84257
[train: 120, 125 / 125] FPS: 21.7 (33.6) , Loss/total: 0.28478 , Loss/sl_clf: -0.01370 , Loss/bbox: 0.04440 , Loss/giou: 0.13412 , bbox_iou: 0.87124 , grad_norm: 49.99839 , reward: -0.00359 , e_mIoU: 0.84017 , mIoU: 0.84376 , mIoU10: 0.84439 , mIoU100: 0.84256
Finished training!Parameter changing curves,
Is there any inference script available for inference on custon video,if it is please share it,Thanks in advance
Hello, I would like to ask about the preprocessing methods for the slt-SiamAttn experiment. Are the preprocessing methods the same as pysot? In other words, can I directly use the dataset that has been preprocessed with pysot? The tracking net dataset is too large; can it be replaced with youtube-bb, imagenet det, or imagenet vid?
Hi, thanks for releasing the code of your great work!
In the paper and code, I notice that you use iou(SLTtracker) - iou(argmax tracker) as per-frame reward and there seems to be no sequence-level reward. Here please correct me if my understanding is wrong. Then I have two questions about this: (1) Since RL could optimize any non-differentiable objectives, why don't directly use the SOT metrics of each sequence to be the reward? (2) Commonly sequence-level RL will use a discount factor gamma to propagate the final reward to each state in the entire sequence. Have you considered to use such discount factors to combine the reward acorss frames?
Thanks again for providing the code of your great work! I hope you could share more thoughts about the setting of reward with me.
Best
After going through this link https://github.com/visionml/pytracking.git , i duplicate the run_video.py when i running it is throwing an error like shown below. will anybody help me from this ,thanks in advance
slt) T7751P1@ubncnd2203bjf:~/Downloads/SLTtrack$ python pytracking/run_video.py slt_transt slt_transt VIDEO_for_sot.mp4 --save_results
Traceback (most recent call last):
File "/home/T7751P1/Downloads/SLTtrack/pytracking/run_video.py", line 38, in
main()
File "/home/T7751P1/Downloads/SLTtrack/pytracking/run_video.py", line 34, in main
run_video(args.tracker_name, args.tracker_param,args.videofile, args.optional_box, args.debug, args.save_results)
File "/home/T7751P1/Downloads/SLTtrack/pytracking/run_video.py", line 20, in run_video
tracker.run_video(videofilepath=videofile, optional_box=optional_box, debug=debug, save_results=save_results)
File "/home/T7751P1/Downloads/SLTtrack/pytracking/../pytracking/evaluation/tracker.py", line 257, in run_video
tracker = MultiObjectWrapper(self.tracker_class, params, self.visdom, fast_load=True)
File "/home/T7751P1/Downloads/SLTtrack/pytracking/../pytracking/evaluation/multi_object_wrapper.py", line 20, in init
self.tracker_copy.initialize_features()
File "/home/T7751P1/Downloads/SLTtrack/pytracking/../pytracking/tracker/slt_transt/slt_transt.py", line 45, in initialize_features
self.params.net.initialize()
File "/home/T7751P1/Downloads/SLTtrack/pytracking/../pytracking/features/net_wrappers.py", line 53, in initialize
super().initialize()
File "/home/T7751P1/Downloads/SLTtrack/pytracking/../pytracking/features/net_wrappers.py", line 37, in initialize
self.load_network()
File "/home/T7751P1/Downloads/SLTtrack/pytracking/../pytracking/features/net_wrappers.py", line 31, in load_network
self.net = load_network(self.net_path, **self.net_kwargs)
File "/home/T7751P1/Downloads/SLTtrack/pytracking/../pytracking/utils/loading.py", line 31, in load_network
net, _ = ltr_loading.load_network(path_full, **kwargs)
File "/home/T7751P1/Downloads/SLTtrack/pytracking/../ltr/admin/loading.py", line 62, in load_network
raise Exception('No matching checkpoint file found')
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
Django
The Web framework for perfectionists with deadlines.
Laravel
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
Microsoft
Open source projects and samples from Microsoft.
Google
Google ❤️ Open Source for everyone.
Alibaba
Alibaba Open Source for everyone
D3
Data-Driven Documents codes.
Tencent
China tencent open source team.