There was a problem training on mot20, but it was normal on mot15

fuyuer233 commented 4 years ago

@ifzhang while train mot20 ，its wrong。

mot/mot20_dla34Traceback (most recent call last):
  File "train.py", line 97, in <module>
    main(opt)
  File "train.py", line 68, in main
    log_dict_train, _ = trainer.train(epoch, train_loader)
  File "/home/user/CHENJIE/O/FairMOT-master/src/lib/trains/base_trainer.py", line 119, in train
    return self.run_epoch('train', epoch, data_loader)
  File "/home/user/CHENJIE/O/FairMOT-master/src/lib/trains/base_trainer.py", line 62, in run_epoch
    for iter_id, batch in enumerate(data_loader):
  File "/home/user/anaconda3/envs/env3.7/lib/python3.7/site-packages/torch/utils/data/dataloader.py", line 819, in __next__
    return self._process_data(data)
  File "/home/user/anaconda3/envs/env3.7/lib/python3.7/site-packages/torch/utils/data/dataloader.py", line 846, in _process_data
    data.reraise()
  File "/home/user/anaconda3/envs/env3.7/lib/python3.7/site-packages/torch/_utils.py", line 369, in reraise
    raise self.exc_type(msg)
IndexError: Caught IndexError in DataLoader worker process 0.
Original Traceback (most recent call last):
  File "/home/user/anaconda3/envs/env3.7/lib/python3.7/site-packages/torch/utils/data/_utils/worker.py", line 178, in _worker_loop
    data = fetcher.fetch(index)
  File "/home/user/anaconda3/envs/env3.7/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/user/anaconda3/envs/env3.7/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/user/CHENJIE/O/FairMOT-master/src/lib/datasets/dataset/jde.py", line 456, in __getitem__
    wh[k] = 1. * w, 1. * h
IndexError: index 128 is out of bounds for axis 0 with size 128

ifzhang commented 4 years ago

When training on MOT20, you should set --K 500. It means the max number of objects in a frame. The original K is 128, but MOT20 has more than 128 objects in a frame, so you should set the max number 500 when training on MOT20.

fuyuer233 commented 4 years ago

ok，thank you

------------------ 原始邮件 ------------------ 发件人: "Yifu Zhang"<notifications@github.com>; 发送时间: 2020年5月4日(星期一) 晚上7:44 收件人: "ifzhang/FairMOT"<FairMOT@noreply.github.com>; 抄送: "Fuyuer。"<1764588497@qq.com>;"Author"<author@noreply.github.com>; 主题: Re: [ifzhang/FairMOT] There was a problem training on mot20, but it was normal on mot15 (#88)

When training on MOT20, you should set --K 500. It means the max number of objects in a frame. The original K is 128, but MOT20 has more than 128 objects in a frame, so you should set the max number 500 when training on MOT20.

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

ifzhang / FairMOT

There was a problem training on mot20, but it was normal on mot15 #88