IndexError: single positional indexer is out-of-bounds

Describe the bug Trying to train a custom object detector when I get the error listed in the title. I think it’s because it’s not reading in my folders with the images and labels? I’ve tried 'labels' ,'/labels' and '/labels/'

Code and Data

from detecto import core, utils, visualize

# Images and XML files in separate folders
dataset = core.Dataset('labels/', 'images/')

image, target = dataset[0]
print(image, target)

model = core.Model(['bat', 'batter', 'pitch', 'field', 'player', 'scoreboard'])

model.fit(dataset)


# Specify the path to your image
image = utils.read_image('images/image0.jpg')
predictions = model.predict(image)

# predictions format: (labels, boxes, scores)
labels, boxes, scores = predictions

print(labels) 
print(boxes)
print(scores)

Stacktrace

Traceback (most recent call last):
  File "c:/Users/julis/Documents/ap-cricket/functions/train.py", line 9, in <module>
    image, target = dataset[0]
  File "C:\Users\julis\AppData\Local\Programs\Python\Python38\lib\site-packages\detecto\core.py", line 148, in __getitem__
    img_name = os.path.join(self._root_dir, self._csv.iloc[idx, 0])
  File "C:\Users\julis\AppData\Local\Programs\Python\Python38\lib\site-packages\pandas\core\indexing.py", line 873, in __getitem__
    return self._getitem_tuple(key)
  File "C:\Users\julis\AppData\Local\Programs\Python\Python38\lib\site-packages\pandas\core\indexing.py", line 1443, in _getitem_tuple
    self._has_valid_tuple(tup)
  File "C:\Users\julis\AppData\Local\Programs\Python\Python38\lib\site-packages\pandas\core\indexing.py", line 702, in _has_valid_tuple
    self._validate_key(k, i)
  File "C:\Users\julis\AppData\Local\Programs\Python\Python38\lib\site-packages\pandas\core\indexing.py", line 1352, in _validate_key
    self._validate_integer(key, axis)
  File "C:\Users\julis\AppData\Local\Programs\Python\Python38\lib\site-packages\pandas\core\indexing.py", line 1437, in _validate_integer
    raise IndexError("single positional indexer is out-of-bounds")
IndexError: single positional indexer is out-of-bounds

Environment:

• OS: Windows 10
• Python version: 3.8
• Detecto version:
• torch version: 1.5.0
• torchvision version : 0.6.0

Additional context Image name is : ‘image0.jpg’ Label name is: ‘image0.xml’

[bgriffen]

I’m also caught up at the moment but will take a look when I can too!

[bgriffen]

This has come up again in another revisit of detecto – Out of curiosity @alankbi, is there a way to not make the strict requirement on image_id having to be both sequential and starting at 0 in both the training and test data? I would have thought after train_test_split, you should be able to just give both dataframes to DataLoader but it seems I then have to reassign all of the image_ids to satisfy that requirement. Presumably, the image_ids can just be abstracted away and just be created on the fly based on the filename column in each train/test dataframe going to the DataLoader. e.g. just like this here. i.e.

df['image_id'] = df.groupby(['filename']).ngroup()

I’m likely missing something behind the scenes that doesn’t allow this, though. On my most recent issue relating to this (I think)…

Begin iterating over training dataset
  0%|                                                                                                        | 0/20 [00:00<?, ?it/s]
---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
<ipython-input-1-0d2a0526aafd> in <module>
      7       d.df[column] = d.df[column].astype('int32')
      8 
----> 9 d.train()

myprogram.py in train(self, epochs)
    146 
    147     def train(self,epochs=10):
--> 148         self.losses = self.model.fit(self.train_loader,self.test_loader,epochs=epochs,verbose=True)
    149 
    150     def save(self):

~/anaconda3/lib/python3.8/site-packages/detecto-1.2.1-py3.8.egg/detecto/core.py in fit(self, dataset, val_dataset, epochs, learning_rate, momentum, weight_decay, gamma, lr_step_size, verbose)
    516 
    517             iterable = tqdm(dataset, position=0, leave=True) if verbose else dataset
--> 518             for images, targets in iterable:
    519                 self._convert_to_int_labels(targets)
    520                 images, targets = self._to_device(images, targets)

~/anaconda3/lib/python3.8/site-packages/tqdm/std.py in __iter__(self)
   1176 
   1177         try:
-> 1178             for obj in iterable:
   1179                 yield obj
   1180                 # Update and possibly print the progressbar.

~/anaconda3/lib/python3.8/site-packages/torch/utils/data/dataloader.py in __next__(self)
    519             if self._sampler_iter is None:
    520                 self._reset()
--> 521             data = self._next_data()
    522             self._num_yielded += 1
    523             if self._dataset_kind == _DatasetKind.Iterable and \

~/anaconda3/lib/python3.8/site-packages/torch/utils/data/dataloader.py in _next_data(self)
    559     def _next_data(self):
    560         index = self._next_index()  # may raise StopIteration
--> 561         data = self._dataset_fetcher.fetch(index)  # may raise StopIteration
    562         if self._pin_memory:
    563             data = _utils.pin_memory.pin_memory(data)

~/anaconda3/lib/python3.8/site-packages/torch/utils/data/_utils/fetch.py in fetch(self, possibly_batched_index)
     42     def fetch(self, possibly_batched_index):
     43         if self.auto_collation:
---> 44             data = [self.dataset[idx] for idx in possibly_batched_index]
     45         else:
     46             data = self.dataset[possibly_batched_index]

~/anaconda3/lib/python3.8/site-packages/torch/utils/data/_utils/fetch.py in <listcomp>(.0)
     42     def fetch(self, possibly_batched_index):
     43         if self.auto_collation:
---> 44             data = [self.dataset[idx] for idx in possibly_batched_index]
     45         else:
     46             data = self.dataset[possibly_batched_index]

~/anaconda3/lib/python3.8/site-packages/detecto-1.2.1-py3.8.egg/detecto/core.py in __getitem__(self, idx)
    151         object_entries = self._csv.loc[self._csv['image_id'] == idx]
    152 
--> 153         img_name = os.path.join(self._root_dir, object_entries.iloc[0, 0])
    154         image = read_image(img_name)
    155 

~/anaconda3/lib/python3.8/posixpath.py in join(a, *p)
     88                 path += sep + b
     89     except (TypeError, AttributeError, BytesWarning):
---> 90         genericpath._check_arg_types('join', a, *p)
     91         raise
     92     return path

~/anaconda3/lib/python3.8/genericpath.py in _check_arg_types(funcname, *args)
    150             hasbytes = True
    151         else:
--> 152             raise TypeError(f'{funcname}() argument must be str, bytes, or '
    153                             f'os.PathLike object, not {s.__class__.__name__!r}') from None
    154     if hasstr and hasbytes:

TypeError: join() argument must be str, bytes, or os.PathLike object, not 'int64

test.csv

width	height	cluster	xmin	ymin	xmax	ymax	image_id	filename
53	133	1	453	274	505	407	0	Position001.jpg
410	189	0	145	238	555	427	1	Position046.jpg
62	127	0	444	273	506	400	1	Position046.jpg
…

and train.csv

width	height	cluster	xmin	ymin	xmax	ymax	image_id	filename
53	133	1	453	274	505	407	0	Position001.jpg
410	189	0	145	238	555	427	1	Position046.jpg
62	127	0	444	273	506	400	1	Position046.jpg
56	123	0	200	265	256	388	1	Position046.jpg
413	192	0	148	226	562	418	2	Position028.jpg
…

As a check…

In [2]: d.train_dataset
Out[2]: <detecto.core.Dataset at 0x7f2af6a5b5e0>

In [3]: d.train_dataset[0]
---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
<ipython-input-3-c58fa18afada> in <module>
----> 1 d.train_dataset[0]

~/anaconda3/lib/python3.8/site-packages/detecto-1.2.1-py3.8.egg/detecto/core.py in __getitem__(self, idx)
    151         object_entries = self._csv.loc[self._csv['image_id'] == idx]
    152 
--> 153         img_name = os.path.join(self._root_dir, object_entries.iloc[0, 0])
    154         image = read_image(img_name)
    155 

~/anaconda3/lib/python3.8/posixpath.py in join(a, *p)
     88                 path += sep + b
     89     except (TypeError, AttributeError, BytesWarning):
---> 90         genericpath._check_arg_types('join', a, *p)
     91         raise
     92     return path

~/anaconda3/lib/python3.8/genericpath.py in _check_arg_types(funcname, *args)
    150             hasbytes = True
    151         else:
--> 152             raise TypeError(f'{funcname}() argument must be str, bytes, or '
    153                             f'os.PathLike object, not {s.__class__.__name__!r}') from None
    154     if hasstr and hasbytes:

TypeError: join() argument must be str, bytes, or os.PathLike object, not 'int64'

These being assigned here in my program…

        self.train_dataset = Dataset(self.fout_train,transform=self.default_trans)
        self.test_dataset = Dataset(self.fout_test) 
        
        self.train_loader = DataLoader(self.train_dataset, batch_size=batch_size, shuffle=True)
        self.test_loader = DataLoader(self.test_dataset, shuffle=True)

Update It comes down to iloc indexing. So when the comments/docs link say the csv file contains…

CSV file contains: filename, width, height, class, xmin, ymin, xmax, ymax

The actual order of those columns matters because of

--> 153 img_name = os.path.join(self._root_dir, object_entries.iloc[0, 0]) in core.py

General comment though - shouldn’t this just be object_entries.loc[0, 'filename'] so it’s agnostic to what order the columns are put in. Also, curious as to people’s thoughts on the nrgoup() function being run to just generate the image_ids in the backend?