index1 = lcnn_line_map[idx, 0] IndexError: index 0 is out of bounds for axis 1 with size 0

[aquexexl]

Hi,
I 've create my own dataset in wireframe format but made a preprocessed with wireframe.py in L-CNN github. There is no error during the preprocessed but when I tried to train with Hawpv3, I had this issue :

Traceback (most recent call last):
File "/opt/conda/lib/python3.9/runpy.py", line 197, in _run_module_as_main
return _run_code(code, main_globals, None,
File "/opt/conda/lib/python3.9/runpy.py", line 87, in _run_code
exec(code, run_globals)
File "/project_ghent/HAWP/hawp/ssl/train.py", line 233, in
main()
File "/project_ghent/HAWP/hawp/ssl/train.py", line 145, in main
train(model, train_loader, optimizer, scheduler, loss_reducer, arguments, output_dir)
File "/project_ghent/HAWP/hawp/ssl/train.py", line 170, in train
for it, data in enumerate(train_loader):
File "/opt/conda/lib/python3.9/site-packages/torch/utils/data/dataloader.py", line 521, in next
data = self._next_data()
File "/opt/conda/lib/python3.9/site-packages/torch/utils/data/dataloader.py", line 1203, in _next_data
return self._process_data(data)
File "/opt/conda/lib/python3.9/site-packages/torch/utils/data/dataloader.py", line 1229, in _process_data
data.reraise()
File "/opt/conda/lib/python3.9/site-packages/torch/_utils.py", line 425, in reraise
raise self.exc_type(msg)
IndexError: Caught IndexError in DataLoader worker process 0.
Original Traceback (most recent call last):
File "/opt/conda/lib/python3.9/site-packages/torch/utils/data/_utils/worker.py", line 287, in _worker_loop
data = fetcher.fetch(index)
File "/opt/conda/lib/python3.9/site-packages/torch/utils/data/_utils/fetch.py", line 44, in fetch
data = [self.dataset[idx] for idx in possibly_batched_index]
File "/opt/conda/lib/python3.9/site-packages/torch/utils/data/_utils/fetch.py", line 44, in
data = [self.dataset[idx] for idx in possibly_batched_index]
File "/project_ghent/HAWP/hawp/ssl/datasets/wireframe_dataset.py", line 1341, in getitem
data = self.train_preprocessing(data)
File "/project_ghent/HAWP/hawp/ssl/datasets/wireframe_dataset.py", line 871, in train_preprocessing
line_map_neg = self.convert_line_map(line_neg, num_junctions)
File "/project_ghent/HAWP/hawp/ssl/datasets/wireframe_dataset.py", line 667, in convert_line_map
index1 = lcnn_line_map[idx, 0]
IndexError: index 0 is out of bounds for axis 1 with size 0

I saw there was empty Lneg (cause only one line in a picture for the train maybe) I don't know if it came from this variable. What do you suggest ?

Thank in advance !

[aquexexl]

The problem came from the shape of Lneg but I don't understand why
Example from my train:
lneg (120, 2, 3)
Lneg (120, 0, 3)
lpos (22, 2, 3)
Lpos (22, 2)

The original code from wireframe.py:
for i0, i1 in combinations(range(len(junc)), 2):
if frozenset([i0, i1]) not in lineset:
v0, v1 = junc[i0], junc[i1]
vint0, vint1 = to_int(v0[:2] / 2), to_int(v1[:2] / 2) #
rr, cc, value = skimage.draw.line_aa(*vint0, *vint1)
lneg.append([v0, v1, i0, i1, np.average(np.minimum(value, llmap[rr, cc]))])
assert len(lneg)!=0
lneg.sort(key=lambda l: -l[-1])
lneg = np.array([l[:2] for l in lneg[:2000]], dtype=np.float32)
Lneg = np.array([l[2:4] for l in lneg][:4000], dtype=int)
print('lneg',np.shape(lneg))
print('Lneg',np.shape(Lneg))
junc = np.array(junc, dtype=np.float32)
Lpos = np.array(lnid, dtype=int)
print('lpos',np.shape(lpos))
print('Lpos',np.shape(Lpos))
lpos = np.array(lpos, dtype=np.float32)
Have you got a solution ?

[cherubicXN]

Hi @aquexexl, the data format of HAWPv3 would be different from the HAWPv2 because I wanted to avoid any mistakes in using human-labeled annotations for self-supervised learning. So, if you want to use the HAWPv3 model in the v2 pipeline, a better way should be to copy the model files to the v2 directory.

For the negative example used in L-CNN, it was used in our HAWPv1 (the CVPR version). But for HAWPv2 and HAWPv3, the negative examples were not used. You can safely set an empty or a dummy negative example to adapt the data format.

Sorry for the late reply, but I am suffering from a difficult time in my life. Hope my reply can provide some information for you.

[aquexexl]

Thank you for your answer,
I solved the problem with Lneg, but in your trainv3, it expected line_map in the tensor, but I only have line_map_pos or line_map_neg. Must I force line_map_pos ?
Traceback (most recent call last):
File "/opt/conda/lib/python3.9/runpy.py", line 197, in _run_module_as_main
return _run_code(code, main_globals, None,
File "/opt/conda/lib/python3.9/runpy.py", line 87, in _run_code
exec(code, run_globals)
File "/project_ghent/HAWP/hawp/ssl/train.py", line 232, in
main()
File "/project_ghent/HAWP/hawp/ssl/train.py", line 145, in main
train(model, train_loader, optimizer, scheduler, loss_reducer, arguments, output_dir)
File "/project_ghent/HAWP/hawp/ssl/train.py", line 178, in train
loss_dict, extra_info = model(images,annotations)
File "/opt/conda/lib/python3.9/site-packages/torch/nn/modules/module.py", line 1051, in _call_impl
return forward_call(*input, **kwargs)
File "/project_ghent/HAWP/hawp/ssl/models/detector.py", line 358, in forward
return self.forward_train(images, annotations=annotations)
File "/project_ghent/HAWP/hawp/ssl/models/detector.py", line 365, in forward_train
targets , metas = self.hafm_encoder(annotations)
File "/project_ghent/HAWP/hawp/ssl/models/hafm.py", line 24, in call
edge_indices = annotations['line_map'][batch_id].triu().nonzero()
KeyError: 'line_map'