Py-faster-rcnn: AssetionError: training py-faster-rcnn with one class

[yawadugyamfi]

I am using py-faster-rcnn to train the 2007 VOC dataset. When I run on the original 21 classes, it works fine, but when I change the original number of classes, I get an error..
I am doing this for 1 class ('person'). I made the following changes to the original codes:

1. file: fcn_root/models/VGG16/faster_rcnn_alt_opt.pt
   layer {
   name: "cls_score"
   type: "InnerProduct"
   bottom: "fc7"
   top: "cls_score"
   inner_product_param {
   num_output: 2
   }
   }

layer {
name: "bbox_pred"
type: "InnerProduct"
bottom: "fc7"
top: "bbox_pred"
inner_product_param {
num_output: 8
}
}

1. file2: fcn_root/lib/datasets/pascal_voc.py

line 28:self._classes = ('__background__', # always index 0
'person')
# removed -1's from the lines below

x1 = float(get_data_from_tag(obj, 'xmin')) y1 = float(get_data_from_tag(obj, 'ymin')) x2 = float(get_data_from_tag(obj, 'xmax')) y2 = float(get_data_from_tag(obj, 'ymax'))

1. file3: ../VOCdevkit/VOCcode/VOCinit.m

VOCopts.classes={...
'person'};

I get the following error after I run this command,

./experiments/scripts/faster_rcnn_alt_opt.sh 3 VGG16 --set EXP_DIR foobarx

I have also attached a full trace of my program running.
Error:
Process Process-1:
Traceback (most recent call last):
File "/usr/lib/python2.7/multiprocessing/process.py", line 258, in _bootstrap
self.run()
File "/usr/lib/python2.7/multiprocessing/process.py", line 114, in run
self._target(_self._args, *_self._kwargs)
File "./tools/train_faster_rcnn_alt_opt.py", line 123, in train_rpn
roidb, imdb = get_roidb(imdb_name)
File "./tools/train_faster_rcnn_alt_opt.py", line 68, in get_roidb
roidb = get_training_roidb(imdb)
File "/home/microway/test/pytest/py-faster-rcnn/tools/../lib/fast_rcnn/train.py", line 121, in get_training_roidb
imdb.append_flipped_images()
File "/home/microway/test/pytest/py-faster-rcnn/tools/../lib/datasets/imdb.py", line 108, in append_flipped_images
assert (boxes[:, 2] >= boxes[:, 0]).all()
AssertionError

error_attachment.txt

[marphil]

Hi,
I have also encountered this error. Seems like something is wrong in at least one of the annotations, an artificial solution is the following: in imdb.py

for b in range(len(boxes)):     
                if boxes[b][2] < boxes[b][0]:
                   boxes[b][0] = 0

assert (boxes[:, 2] >= boxes[:, 0]).all()

Hope this helps!

[moyans]

@marphil I get the following error:
Process Process-1:
Traceback (most recent call last):
File "/usr/lib/python2.7/multiprocessing/process.py", line 258, in _bootstrap
self.run()
File "/usr/lib/python2.7/multiprocessing/process.py", line 114, in run
self._target(_self._args, *_self._kwargs)
File "./tools/train_faster_rcnn_alt_opt.py", line 122, in train_rpn
roidb, imdb = get_roidb(imdb_name)
File "./tools/train_faster_rcnn_alt_opt.py", line 67, in get_roidb
roidb = get_training_roidb(imdb)
File "/media/dl/c0d34b51-8a78-4364-bc19-35003bf01885/RCNN/py-faster-rcnn/tools/../lib/fast_rcnn/train.py", line 118, in get_training_roidb
imdb.append_flipped_images()
File "/media/dl/c0d34b51-8a78-4364-bc19-35003bf01885/RCNN/py-faster-rcnn/tools/../lib/datasets/imdb.py", line 106, in append_flipped_images
boxes = self.roidb[i]['boxes'].copy()
File "/media/dl/c0d34b51-8a78-4364-bc19-35003bf01885/RCNN/py-faster-rcnn/tools/../lib/datasets/imdb.py", line 67, in roidb
self._roidb = self.roidb_handler()
File "/media/dl/c0d34b51-8a78-4364-bc19-35003bf01885/RCNN/py-faster-rcnn/tools/../lib/datasets/pascal_voc.py", line 103, in gt_roidb
for index in self.image_index]
File "/media/dl/c0d34b51-8a78-4364-bc19-35003bf01885/RCNN/py-faster-rcnn/tools/../lib/datasets/pascal_voc.py", line 204, in _load_pascal_annotation
x1 = float(bbox.find('xmin').text)
ValueError: invalid literal for float(): 737,

[DiegoPortoJaccottet]

Hello yawadugyamfi, did you find a solution to your error? Would you share it with us please?

Never mind, I found the answer. There should be no <ymin>0</ymin> nor <xmin>0</xmin> in an object.

I changed that in the Annotations folder with the two commands (./ means the current folder is Annotations):
grep -rl '<ymin>0</ymin>' ./ | xargs sed -i 's#<ymin>0</ymin>#<ymin>1</ymin>#g'
and
grep -rl '<xmin>0</xmin>' ./ | xargs sed -i 's#<xmin>0</xmin>#<xmin>1</xmin>#g'

Now training is running smoothly. .

[aragon111]

@marphil in which function you wrote those lines?

[aragon111]

@DiegoPortoJaccottet I got sed: no input files

[alviur]

@marphil that worked for me!.

[CWOA]

@DiegoPortoJaccottet your fix worked for me, thanks! The @marphil fix produces the same assertion error in my case.

[harora]

@DiegoPortoJaccottet Even i got sed : no input files. I have my annotation files in that folder ( *.xml)

[DiegoPortoJaccottet]

@harora
You got 'sed: no input files' from the commands I described or from Faster-RCNN?

[XiangyuChu]

@moyans I get the same error with you.

y2= float(bbox.find('ymax').text)
ValueError: invalid literal for float(): 196006001002176.jpg

Have you found any solution to it? Thank you

[moyans]

you need check this img and label.

来自 魅族 MX5

-------- 原始邮件 --------
发件人：Xiang [e-mail]
时间：周五 5月5日 09:52
收件人：rbgirshick/py-faster-rcnn [e-mail]
抄送：moyans [e-mail],Mention [e-mail]
主题：Re: [rbgirshick/py-faster-rcnn] AssetionError: training py-faster-rcnn with one class (#34)

@moyans I get the same error with you.

y2= float(bbox.find('ymax').text)
ValueError: invalid literal for float(): 196006001002176.jpg

Do you find any solution to it? Thank you

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

{"api_version":"1.0","publisher":{"api_key":"05dde50f1d1a384dd78767c55493e4bb","name":"GitHub"},"entity":{"external_key":"github/rbgirshick/py-faster-rcnn","title":"rbgirshick/py-faster-rcnn","subtitle":"GitHub repository","main_image_url":"https://cloud.githubusercontent.com/assets/143418/17495839/a5054eac-5d88-11e6-95fc-7290892c7bb5.png","avatar_image_url":"https://cloud.githubusercontent.com/assets/143418/15842166/7c72db34-2c0b-11e6-9aed-b52498112777.png","action":{"name":"Open in GitHub","url":"https://github.com/rbgirshick/py-faster-rcnn"}},"updates":{"snippets":[{"icon":"PERSON","message":"@XiangyuChu in #34: @moyans I get the same error with you. rnrny2= float(bbox.find('ymax').text)rnValueError: invalid literal for float(): 196006001002176.jpgrnrnDo you find any solution to it? Thank you "}],"action":{"name":"View Issue","url":"https://github.com/rbgirshick/py-faster-rcnn/issues/34#issuecomment-299353335"}}}

[XiangyuChu]

@moyans 196006001002176.jpg is very odd. My images are usually "000000.jpg", same as VOC2007.
And I have removed "-1" from the lines below
x1 = float(bbox.find( 'xmin').text)
y1 = float(bbox.find('ymin').text)
x2 = float(bbox.find('xmax').text)
y2 = float(bbox.find('ymax').text)

[moyans]

check the label out of range?

来自 魅族 MX5

-------- 原始邮件 --------
发件人：Xiang [e-mail]
时间：周五 5月5日 10:32
收件人：rbgirshick/py-faster-rcnn [e-mail]
抄送：moyans [e-mail],Mention [e-mail]
主题：Re: [rbgirshick/py-faster-rcnn] AssetionError: training py-faster-rcnn with one class (#34)

@moyans 196006001002176.jpg is very odd. My images are usually "000000.jpg", same as VOC2007.
And I have removed "-1" from the lines below
x1 = float(get_data_from_tag(obj, 'xmin'))
y1 = float(get_data_from_tag(obj, 'ymin'))
x2 = float(get_data_from_tag(obj, 'xmax'))
y2 = float(get_data_from_tag(obj, 'ymax'))

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

{"api_version":"1.0","publisher":{"api_key":"05dde50f1d1a384dd78767c55493e4bb","name":"GitHub"},"entity":{"external_key":"github/rbgirshick/py-faster-rcnn","title":"rbgirshick/py-faster-rcnn","subtitle":"GitHub repository","main_image_url":"https://cloud.githubusercontent.com/assets/143418/17495839/a5054eac-5d88-11e6-95fc-7290892c7bb5.png","avatar_image_url":"https://cloud.githubusercontent.com/assets/143418/15842166/7c72db34-2c0b-11e6-9aed-b52498112777.png","action":{"name":"Open in GitHub","url":"https://github.com/rbgirshick/py-faster-rcnn"}},"updates":{"snippets":[{"icon":"PERSON","message":"@XiangyuChu in #34: @moyans 196006001002176.jpg is very odd. My images are usually "000000.jpg", same as VOC2007. rnAnd I have removed "-1" from the lines belowrn x1 = float(get_data_from_tag(obj, 'xmin'))rn y1 = float(get_data_from_tag(obj, 'ymin'))rn x2 = float(get_data_from_tag(obj, 'xmax')) rn y2 = float(get_data_from_tag(obj, 'ymax'))"}],"action":{"name":"View Issue","url":"https://github.com/rbgirshick/py-faster-rcnn/issues/34#issuecomment-299357112"}}}

[firestonelib]

in the Annotation folder I run
grep -rl '<ymin>0</ymin>' ./ | xargs sed -i 's#<ymin>0</ymin>#<ymin>1</ymin>#g'
I got: sed : no input files just like @harora . I want to know why. @DiegoPortoJaccottet

[XiangyuChu]

@moyans Thanks for your help. I have solved the error. I found an error in "output.txt", i.e. 002175.jpg c 669 155 769 **196006001002176.jpg p 207 188 229 257. What puzzles me is that I cannot find error in .xml. Anyway, it is solved. Thank you again!

[iamlxb3]

I replaced xmin 0 and ymin 0 with xmin 1 and ymin 1, all things resolved!

see more https://github.com/rbgirshick/py-faster-rcnn/issues/34

[ankitcsr]

in the Annotation folder I run
grep -rl '<ymin>0</ymin>' ./ | xargs sed -i 's#<ymin>0</ymin>#<ymin>1</ymin>#g'
I got: sed : no input files just like @harora . I want to know why. @DiegoPortoJaccottet

replace 0 with 0.0