mhruz/potr

Pose Transformer

potr's Issues

Task: parametry v yaml

• zapsat do config yamlu vsechny parametry s defaultnimi hodnotami a vyzkouset

Bug: matcher.py

POTR/deformable_potr_plus/models/matcher.py

Line 89 in 7b8c12a

indices = [linear_sum_assignment(c[i]) for i, c in enumerate(C.split(sizes, -1))]

poustel jsem depotr plus s efficientnetv2_rw_s...bezelo to, ale obe
instance padli po nekolika epochach s tou samou chybou
(hpoes\experiments\depotr_plus\exp10 a exp11):

Traceback (most recent call last):
File
"/storage/plzen1/home/jkanis/.conda/envs/deformable_potr/lib/python3.7/runpy.py",
line 193, in _run_module_as_main
"main", mod_spec)
File
"/storage/plzen1/home/jkanis/.conda/envs/deformable_potr/lib/python3.7/runpy.py",
line 85, in _run_code
exec(code, run_globals)
File
"/auto/plzen4-ntis/projects/cv/hpoes/experiments/depotr_plus/exp10/deformable_potr_plus/main.py",
line 344, in
main(args)
File
"/auto/plzen4-ntis/projects/cv/hpoes/experiments/depotr_plus/exp10/deformable_potr_plus/main.py",
line 286, in main
train_stats = train_one_epoch(model, criterion, data_loader_train,
optimizer, device, epoch, log_wandb=bool(args.wandb_key))
File
"/auto/plzen4-ntis/projects/cv/hpoes/experiments/depotr_plus/exp10/deformable_potr_plus/engine.py",
line 39, in train_one_epoch
loss_dict = criterion(outputs, targets)
File
"/storage/plzen1/home/jkanis/.conda/envs/deformable_potr/lib/python3.7/site-packages/torch/nn/modules/module.py",
line 727, in _call_impl
result = self.forward(*input, **kwargs)
File
"/auto/plzen4-ntis/projects/cv/hpoes/experiments/depotr_plus/exp10/deformable_potr_plus/models/deformable_potr_plus.py",
line 319, in forward
indices = self.matcher(outputs, targets)
File
"/storage/plzen1/home/jkanis/.conda/envs/deformable_potr/lib/python3.7/site-packages/torch/nn/modules/module.py",
line 727, in _call_impl
result = self.forward(*input, **kwargs)
File
"/auto/plzen4-ntis/projects/cv/hpoes/experiments/depotr_plus/exp10/deformable_potr_plus/models/matcher.py",
line 89, in forward
indices = [linear_sum_assignment(c[i]) for i, c in
enumerate(C.split(sizes, -1))]
File
"/auto/plzen4-ntis/projects/cv/hpoes/experiments/depotr_plus/exp10/deformable_potr_plus/models/matcher.py",
line 89, in
indices = [linear_sum_assignment(c[i]) for i, c in
enumerate(C.split(sizes, -1))]
File
"/storage/plzen1/home/jkanis/.conda/envs/deformable_potr/lib/python3.7/site-packages/scipy/optimize/_lsap.py",
line 101, in linear_sum_assignment
a, b = _lsap_module.calculate_assignment(cost_matrix.T)
ValueError: matrix contains invalid numeric entries

...podle Zdenka to vypada, ze se tam do matice dostane naky
nekonecno....tak by to chtelo asi osetrit...treba nakym ifem....

train and test?

Hello, you have done a great job, I want to reproduce your project, but the strength is limited, can you provide your training and test commands? Looking forward to hearing from you！

'util' module

Hi, I'm Daeyun.

First, thank you for providing code.
I just clone and implement your code as you comment on README.md

then the error occurs on command in terminal

Command
python crops2h5.py ./nyu/train_comrefV2V_3Dproj.json
Error
No module named 'util'

how can I fix this?

How to use this code to NYU/icvl hands dataset?

i read the paper(MuTr: Multi-Stage Transformer for Hand Pose Estimation from Full-Scene Depth Image), and got this code， but i found it difficult to use the code to NYU/icvl dataset, may i ask if there has README.md about how to apply the model in hand pose dataset?

Feature: wandb

• pridat jmeno uzivatele a projektu ve wandb do parametru - ted je to v kodu
• pridat pojmenovani do wandy do parametru - ted se to pojmenuje nahodne ve wandb
• pridat train loss za celou epochu do wandb
• volat inicializaci wandb az je treba neco zapsat, jinak se vytvari zaznam ve wandb i kdyz program spadne napr. kvuli spatnemu parametru