KeyError: 'Nf1aqv5Fg5o_0'

[meimeixu520]

Hi,thanks your excellent work. When I train ATOM using the setting 'atom_paper.py',it uses the dataset TrackingNet,then shows the KeyError,but the dataset is correct.Can you tell me why?Look foward your reply! Best regards!

[meimeixu520]

[meimeixu520]

the error is as follows:

Training: bbreg atom_paper Traceback (most recent call last): File "run_training.py", line 56, in main() File "run_training.py", line 51, in main run_training(args.train_module, args.train_name, args.cudnn_benchmark) File "run_training.py", line 40, in run_training expr_func(settings) File "../ltr/train_settings/bbreg/atom_paper.py", line 27, in run trackingnet_train = TrackingNet(settings.env.trackingnet_dir, set_ids=list(range(11))) File "../ltr/dataset/tracking_net.py", line 71, in init self.seq_to_class_map, self.seq_per_class = self._load_class_info() File "../ltr/dataset/tracking_net.py", line 86, in _load_class_info class_name = seq_to_class_map[seq[1]] KeyError: 'Nf1aqv5Fg5o_0'

[goutamgmb]

Hi,

Thanks for notifying the issue. I am not entirely sure why the class name is missing for that sequence. Anyways I have pushed a fix for the issue. Could you try again with the latest commit?

[meimeixu520]

Great，it works!But I account for another error,for example: I check the parameters ,it is correct,so I don't know why it is wrong.Can you tell me? Best regards!

[goutamgmb]

It was another bug introduced in the latest release :(. I have fixed it now. Can you check again?

[meimeixu520]

OK,it starts trainning!Thanks for your reply! I have another question,in the tracking prase,pytracking/pytracking/tracker/base/basetracker.py,it changes a lot,there is no previous initialization process.So how does it initialize?

[meimeixu520]

when it run through the seventh epoch,it produces the following error:

[goutamgmb]

Regarding the first question, I am not sure what do you mean. Could you please clarify?

Regarding the "too small bounding box" issue, yes we have seen this issue before. We will try to fix it.

[danier97]

Hi,

The commit to fix this issue seems to have missed out other parts of the code that can raise the key error. E.g. in the same file (tracking_net.py) line 126 in function _get_class:

def _get_class(self, seq_id):
    seq_name = self.sequence_list[seq_id][1]
    return self.seq_to_class_map[seq_name]

Thanks

[lerohiso]

you can fix the "KeyError" by adding "Nf1aqv5Fg5o_0 airplane" line in trackingnet_classmap.txt file

[meimeixu520]

Thank your reply,can you add my wechat,19966519787,then can you send me the picture or the way of solving the problem!Thanks very much! Best wishes!

------------------ 原始邮件 ------------------ 发件人: "visionml/pytracking" <notifications@github.com>; 发送时间: 2020年9月10日(星期四) 下午3:38 收件人: "visionml/pytracking"<pytracking@noreply.github.com>; 抄送: "别做白日梦了??"<1398629150@qq.com>;"Author"<author@noreply.github.com>; 主题: Re: [visionml/pytracking] KeyError: 'Nf1aqv5Fg5o_0' (#156)

you can fix the "KeyError" by adding "Nf1aqv5Fg5o_0 airplane" line in trackingnet_classmap.txt file

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or unsubscribe.

[martin-danelljan]

@lerohiso Thank you very much for giving the solution! It should be fixed now.

[zj717754140]

I clone the repo recently,I also meet the same problem KeyError: 'Nf1aqv5Fg5o_0' in Training.How to solve this

KeyError: Caught KeyError in DataLoader worker process 0. Original Traceback (most recent call last): File "/home/jzhang/anaconda3/envs/pytracking/lib/python3.7/site-packages/torch/utils/data/_utils/worker.py", line 185, in _worker_loop data = fetcher.fetch(index) File "/home/jzhang/anaconda3/envs/pytracking/lib/python3.7/site-packages/torch/utils/data/_utils/fetch.py", line 44, in fetch data = [self.dataset[idx] for idx in possibly_batched_index] File "/home/jzhang/anaconda3/envs/pytracking/lib/python3.7/site-packages/torch/utils/data/_utils/fetch.py", line 44, in <listcomp> data = [self.dataset[idx] for idx in possibly_batched_index] File "/home/jzhang/VisualTracking/pytracking/ltr/data/sampler.py", line 159, in __getitem__ train_frames, train_anno, meta_obj_train = dataset.get_frames(seq_id, train_frame_ids, seq_info_dict) File "/home/jzhang/VisualTracking/pytracking/ltr/dataset/tracking_net.py", line 143, in get_frames obj_class = self._get_class(seq_id) File "/home/jzhang/VisualTracking/pytracking/ltr/dataset/tracking_net.py", line 126, in _get_class return self.seq_to_class_map[seq_name] KeyError: 'Nf1aqv5Fg5o_0'

[martin-danelljan]

@zj717754140 When was "recently" exactly? Can you try pulling first?