about the wider face data

Question

about the wider face data

Usernamezhx opened this issue 8 years ago · 23 comments

hi
I follow your guide to start this project. I download the data WIDER face.but when i run the command:"No dataset given" , i find that the faster_rcnn_end2end.sh scripts only support only voc and coco data. and then i try to convert the wider face data to the voc style . but i meet a new question:assert (boxes[:, 2] >= boxes[:, 0]).all() .i try all i can do ,but it does not work.

Answer 1 · 2017-02-20T20:43:19.000Z

The faster_rcnn_end2end.sh script does support the wider dataset. You might need to exactly follow the instructions in the readme file.

Answer 2 · 2017-02-21T02:00:21.000Z

hi playerkk
I am a fresh bird,you said the faster_rcnn_end2end.sh script does support the wider dataset,but i open the faster_rcnn_end2end.sh. it said that :
case $DATASET in
pascal_voc)
TRAIN_IMDB="voc_2007_trainval"
TEST_IMDB="voc_2007_test"
PT_DIR="pascal_voc"
ITERS=70000
;;
coco)
# This is a very long and slow training schedule
# You can probably use fewer iterations and reduce the
# time to the LR drop (set in the solver to 350,000 iterations).
TRAIN_IMDB="coco_2014_train"
TEST_IMDB="coco_2014_minival"
PT_DIR="coco"
ITERS=490000
;;
*)
echo "No dataset given"
exit
;;
esac
i have no idea .i need your help,thank you

Answer 3 · 2017-02-21T02:03:12.000Z

Are you talking about this file? https://github.com/playerkk/face-py-faster-rcnn/blob/master/experiments/scripts/faster_rcnn_end2end.sh

Answer 4 · 2017-02-21T02:13:40.000Z

oh my god ......i use the rbgirshick faster rcnn faster_rcnn_end2end.sh......many many many thanks.......

Answer 5 · 2017-02-21T02:44:40.000Z

i clone your project . thank you very much........ : )

Answer 6 · 2017-02-21T07:33:06.000Z

hi playerkk
I am sorry to bother you .I think i should run it first and then to analysis it.
AssertionError: Path does not exist: data/WIDER/train_face_train_annot.txt .
i think it has some problem with the file name .

image_set_file = self._name + 'face' + self._image_set + '_annot.txt'
I think it should not produce the name :train_face_train_annot.txt

I input the command: **voole@zhx:~/face-py-faster-rcnn$ ./experiments/scripts/faster_rcnn_end2end.sh 0 ZF wider **

thanks for your help
zhanghongxing

Answer 7 · 2017-02-21T10:07:14.000Z

i really really really really really need your help, if you dont have time ,i will be waiting for you.

Answer 8 · 2017-02-21T15:22:08.000Z

I would strongly recommend you to read the instructions carefully before running the cod. https://github.com/playerkk/face-py-faster-rcnn/blob/master/README.md

Answer 9 · 2017-02-22T02:31:35.000Z

I am sorry to bother you again ,I have read the README for several times , http://blog.csdn.net/zhanghongxing007/article/details/56479206
I write down the details, I hope you can read for several minutes, thanks for your help.

Answer 10 · 2017-02-22T02:36:37.000Z

You didn't do the step 4 of the section "prepare the training data".

Answer 11 · 2017-02-22T05:55:45.000Z

first of all . thank you very much. that i use the folder WIDER/WIDER_train/ to Create symlinks. after that , it said that need:
AssertionError: Selective search data not found at: /home/voole/face-py-faster-rcnn/data/WIDER/train.mat
so i down the wider_face_split.zip from the WIDER FACE NET. and then i change the name wider_face_train.mat to train.mat
after that i meet a new problem:
NotImplementedError: Please use HDF reader for matlab v7.3 files
so i have to change the scipy io to fit the v7.3 but................i cant find the loadmat function............

Answer 12 · 2017-02-22T15:20:41.000Z

I tested the repository on another machine. I also had some errors of loading the WIDER dataset. But they are different from yours. I've updated the source files. Please update the repository by doing "git pull origin master" and re-run the program as "./experiments/scripts/faster_rcnn_end2end.sh 0 VGG16 wider"

Sorry for the inconvenience. Let me know if it works.

Answer 13 · 2017-02-23T11:21:48.000Z

hi playerkk
First of all ,thanks for your reply. I am sorry to bother you again again again....but i have be here for one day.
Traceback (most recent call last): File "./tools/train_net.py", line 106, in <module> imdb, roidb = combined_roidb(args.imdb_name) File "./tools/train_net.py", line 71, in combined_roidb roidbs = [get_roidb(s) for s in imdb_names.split('+')] File "./tools/train_net.py", line 68, in get_roidb roidb = get_training_roidb(imdb) File "/home/voole/face-py-faster-rcnn/tools/../lib/fast_rcnn/train.py", line 118, in get_training_roidb imdb.append_flipped_images() File "/home/voole/face-py-faster-rcnn/tools/../lib/datasets/imdb.py", line 106, in append_flipped_images boxes = self.roidb[i]['boxes'].copy() File "/home/voole/face-py-faster-rcnn/tools/../lib/datasets/imdb.py", line 67, in roidb self._roidb = self.roidb_handler() File "/home/voole/face-py-faster-rcnn/tools/../lib/datasets/face.py", line 227, in selective_search_roidb ss_roidb = self._load_selective_search_roidb(gt_roidb) File "/home/voole/face-py-faster-rcnn/tools/../lib/datasets/face.py", line 247, in _load_selective_search_roidb 'Selective search data not found at: {}'.format(filename)
AssertionError: Selective search data not found at: /home/voole/face-py-faster-rcnn/data/WIDER/wider.mat
wider.mat? it means the wider_face_train.mat?
Look forward to your reply

Answer 14 · 2017-02-23T20:52:30.000Z

It's weird. Since we are using RPN to generate proposals, there is no need to call the _load_selective_search_roidb function. Can you post your command to run the code?

Answer 15 · 2017-02-24T02:23:59.000Z

hi playerkk
thanks for your reply. i write down the detail at the : http://blog.csdn.net/zhanghongxing007/article/details/56479206
many thanks.

Answer 16 · 2017-02-24T02:29:28.000Z

Sorry, I should have mentioned only VGG16 is currently supported.

Answer 17 · 2017-02-24T02:46:03.000Z

Hi playerkk
yeah,i use the command:./experiments/scripts/faster_rcnn_end2end.sh 0 VGG16 wider. i use the VGG16 net.

Answer 18 · 2017-02-24T03:07:02.000Z

i have the same quetion .the faster rcnn use the RPN net instead of the selective_search. I called some old script for fast-rcnn ?

Answer 19 · 2017-02-24T03:08:35.000Z

You are training a faster r-cnn. Do you have errors right now?

Answer 20 · 2017-02-24T03:18:30.000Z

when i input the conmmand: './experiments/scripts/faster_rcnn_end2end.sh 0 VGG16 wider'
it still point that : ' Selective search data not found at: /home/voole/face-py-faster-rcnn/data/WIDER/wider.mat'
i am sorry to bother you.

Answer 21 · 2017-02-24T03:20:37.000Z

It seems case sensitive. Try vgg16 instead of VGG16. Also do "git pull origin master" to update the source files.

Answer 22 · 2017-02-24T04:08:51.000Z

You might also need to delete *.pkl files in the cache folder.

Answer 23 · 2017-02-28T07:46:11.000Z

hi man
i have solved this question~~~,because of the easydict, my version is 1.4 .even i config the yml. but easydict did not pass the parameter to the train_net.py. so ...... the solution is to update your easydict. thank you very much very much very much .now it works!