Closed QianyuanLiu closed 6 years ago
InnerPeace-Wu
commented
6 years ago Hi,
Firstly, make sure that you did finish the data preparing steps properly.
There may be something wrong of your roidb. To do further debugging, you print out the value of roidb[i]['image'] before the line, and to see that if the path is valid.
One simple solution maybe deleting the directory of 1.2_cache and re-generate roidb.
Wish it helps.
QianyuanLiu
commented
6 years ago @InnerPeace-Wu
Very thankful for your reply !
1)
I recheck the folder /home/XXX/docment/densecap-tensorflow/git/visual_genome/1.2_cache
there are 154797 files in folder train_gt_roidb and 5001 files in val_gt_roidb
└── visual_genome
├── 1.2
│ ├── test_gt_regions
│ ├── train_gt_regions
│ ├── val_gt_regions
│ └── vocabulary.txt
└── 1.2_cache
├── train_gt_roidb
└── val_gt_roidb
2) I print the value roidb[0]['image'] and get this:
*************************************************************************
im = cv2.imread(roidb[0]['image'])
None
3) I delete the old and re-generate roidb.
XXX@XXX-PC:~/docment/densecap-tensorflow/info$ python read_regions.py --version 1.2 --vg_path
../VG
XXX@XXX-PC:~/docment/densecap-tensorflow/info$ cd ../lib/
XXX@XXX-PC:~/docment/densecap-tensorflow/lib$ python preprocess.py --version 1.2 --path ../VG --output_dir ../git/visual_genome --max_words 10 --limit_ram
split image number: 77398 for split name: train
start loading json files...
0.590941 seconds for loading
train: 100%|███████████████████████████| 108077/108077 [03:08<00:00, 573.80it/s]
processing train set with time: 188.39 seconds
there are 272 invalid bboxes out of 3684063
there are 3 empty phrases after triming
Found 56945 unique word tokens.
Using vocabulary size 10000.
The least frequent word in our vocabulary is 'ruff' and appeared 14 times.
Dumping vocabulary to file: ../git/visual_genome/1.2/vocabulary.txt
Done.
split image number: 5000 for split name: val
start loading json files...
0.452126 seconds for loading
val: 100%|████████████████████████████| 108077/108077 [00:18<00:00, 5692.37it/s]
processing val set with time: 18.99 seconds
there are 14 invalid bboxes out of 237362
there are 0 empty phrases after triming
split image number: 5000 for split name: test
start loading json files...
0.303303 seconds for loading
test: 100%|███████████████████████████| 108077/108077 [00:22<00:00, 4841.91it/s]
processing test set with time: 22.32 seconds
there are 17 invalid bboxes out of 238069
there are 0 empty phrases after trimingwhen i try to train by typing:
sudo CUDA_VISIBLE_DEVICES="0" bash scripts/dense_cap_train.sh
isual_genome_1.2 res50 resnet_v1_50/res50.ckpt git/visual_genome 1the same ERROR and same massages are ouput .
Ckpt path: resnet_v1_50/res50.ckpt
Traceback (most recent call last):
File "./tools/train_net.py", line 214, in <module>
main()
File "./tools/train_net.py", line 210, in main
max_iters=args.max_iters)
File "/home/XXX/docment/densecap-tensorflow/tools/../lib/dense_cap/train.py", line 485, in train_net
sw.train_model(sess, max_iters)
File "/home/XXX/docment/densecap-tensorflow/tools/../lib/dense_cap/train.py", line 356, in train_model
blobs = self.data_layer.forward()
File "/home/XXX/docment/densecap-tensorflow/tools/../lib/fast_rcnn/layer.py", line 99, in forward
blobs = self._get_next_minibatch()
File "/home/XXX/docment/densecap-tensorflow/tools/../lib/fast_rcnn/layer.py", line 95, in _get_next_minibatch
return get_minibatch(minibatch_db)
File "/home/XXX/docment/densecap-tensorflow/tools/../lib/fast_rcnn/minibatch.py", line 41, in get_minibatch
im_blob, im_scales, roidb = _get_image_blob(roidb, random_scale_inds)
File "/home/XXXdocment/densecap-tensorflow/tools/../lib/fast_rcnn/minibatch.py", line 101, in _get_image_blob
cfg.TRAIN.MAX_SIZE)
File "/home/XXX/docment/densecap-tensorflow/tools/../lib/utils/blob.py", line 37, in prep_im_for_blob
im = im.astype(np.float32, copy=False)
AttributeError: 'NoneType' object has no attribute 'astype'Is there my command any wrong ?
other quetion Should i add or modify configurations in root/scripts/dense_cap_config.yml ?
waitting any help :(
thx !
InnerPeace-Wu
commented
6 years ago Please add code
print(roidb[0]['image'])before
im = cv2.imread(roidb[0]['image'])to see the if the path if valid.
QianyuanLiu
commented
6 years ago @InnerPeace-Wu
Hello，I add this code ：print(roidb[0]['image'])in here：
for i in xrange(num_images):
**print(roidb[0]['image']) ##**
im = cv2.imread(roidb[i]['image'])
if roidb[i]['flipped']:
im = im[:, ::-1, :]
target_size = cfg.TRAIN.SCALES[scale_inds[i]]
im, im_scale = prep_im_for_blob(im, cfg.PIXEL_MEANS, target_size,
cfg.TRAIN.MAX_SIZE)
im_scales.append(im_scale)
processed_ims.append(im)and get some output massage like this：
../VG/images/2409215.jpg
and same error info：
AttributeError: 'NoneType' object has no attribute 'astype'
PS: I add this code print (im) following closely theim = cv2.imread(roidb[0]['image'])
last, print like this:
../VG/images/2409215.jpg
<type 'NoneType'>
Noneso, is cv2.read failed ?
InnerPeace-Wu
commented
6 years ago The problem seems about your image path, where you used a relative path as ../VG/images/2409215.jpg (I bet the image id 2409215 is valid, is it?). However you should use a absolute path like /home/user/git/VG/images/id.jpg.
One way to solve this, you should re-preprocessing data:
Firstly, deleting the directory of 1.2_cache.
Secondly, preprocessing the data with code:
$ cd $ROOT/lib
$ python preprocess.py --version [version] --path [raw_data_path] \
--output_dir [dir] --max_words [max_len] --limit_ramNote: the [raw_data_path] should be the absolute path of VG like /home/user/git/VG. Or you can refer to my example code.
For your case, you can easily address the issue by changing the relative path to the absolute one by changing the line to:
path = '/home/XXX/docment/densecap-tensorflow' + roidb[i]['image'][2:]
im = cv2.imread(path)Best,
QianyuanLiu
commented
6 years ago congratulation ! It works ! Thank you ! 👍 👍 👍
Hello, I am a new TF-boy and help some help :) In your tutorial,
when i try to train my model by typing ：
sudo CUDA_VISIBLE_DEVICES="0" bash scripts/dense_cap_train.sh visual_genome_1.2 res50 resnet_v1_50/res50.ckpt git/visual_genome 1I get the ERROR:
There are my sys-info:
ubuntu16.04 GeForce GTX TITAN RAM:8G CUDA:8.0 cudnn:6.0 TF:1.4
I review the code and think the net wasn't feed any data, I locate the issues in this file:
/lib/utils/blob.pyso, i guess the net read nothing from roidb. and how do i fix it ?
PS: the tree of my workspace:
any help will be appreciated. THX : )