Hi,
I want to improve the gender accuracy,
so I'd like to finetune your model (model_epoch_190) with my dataset
(some croped face with 21 landmarks and gender label)

but I still don't know how to start it,
should I add my dataset to aflw.sqlite ?
can you give me some tips ?

Any help will be highly appreciated,Thanks!

subj

Hi @takiyu ,
Thanks for this amazing work.
Are you going to take this further?
Any plans to implement hyperface for multiple faces for much larger images?
I would like to contribute to move it further.
Thanks!!

Trying to find any good env configuration where train.py uses gpu. Tried various chainer and cupy versions. Was able to run train on CPU but thats it. Anybody succeed with GPU ?

Hi,
I look to your entire code. I was wondering if the function called "proposal_region" in your code is the one called FaceRectCalculator present in the HyperFace paper.
If so, how did you understand what it if should perform?

Thanks
Anna

I get an error when I run:

python ./scripts/demo_live.py --model model_epoch_190

I have a MacOsX system and Here is the log:

Traceback (most recent call last):
  File "/anaconda/envs/hyperface/lib/python2.7/multiprocessing/process.py", line 258, in _bootstrap
    self.run()
  File "/anaconda/envs/hyperface/lib/python2.7/multiprocessing/process.py", line 114, in run
    self._target(*self._args, **self._kwargs)
  File "/Users/aminbarekatain/Desktop/hyperface/scripts/extensions/imgliveuploader/app.py", line 99, in new_server
    logger=False, engineio_logger=False)
  File "/anaconda/envs/hyperface/lib/python2.7/site-packages/flask_socketio/__init__.py", line 119, in __init__
    self.init_app(app, **kwargs)
  File "/anaconda/envs/hyperface/lib/python2.7/site-packages/flask_socketio/__init__.py", line 144, in init_app
    self.server = socketio.Server(**self.server_options)
  File "/anaconda/envs/hyperface/lib/python2.7/site-packages/socketio/server.py", line 89, in __init__
    self.eio = self._engineio_server_class()(**engineio_options)
  File "/anaconda/envs/hyperface/lib/python2.7/site-packages/engineio/server.py", line 122, in __init__
    raise ValueError('Invalid async_mode specified')

Can you help me to fix this problem?

Hi @takiyu
I have a problem that must the Test part use gpu?
Thank you very much!

Hello, i am reimplement hyperfaec in pytorch these days.
I meet some problems, and I hope to talk about these questions with you.
May i have your e-mail or other social media ?

Hello, thanks for your impressive work!
When I run your code to reproduce the work, I met the following errors:

[2017/08/28 09:51:03] WARNING (13199) datasets : No selective search rectangle
Process Process-10:...........................................] 4.93%
Traceback (most recent call last):ochs
File "/home/users/yuxi.feng/anaconda2/lib/python2.7/multiprocessing/process.py", line 258, in _bootstrap
self.run()
File "/home/users/yuxi.feng/anaconda2/lib/python2.7/multiprocessing/process.py", line 114, in run
self._target(*self._args, **self._kwargs)
File "/home/users/yuxi.feng/anaconda2/lib/python2.7/site-packages/chainer/iterators/multiprocess_iterator.py", line 162, in _worker
out_queue.put(dataset[index])
File "/home/users/yuxi.feng/anaconda2/lib/python2.7/site-packages/chainer/dataset/dataset_mixin.py", line 30, in getitem
return self.get_example(index)
File "/home/users/yuxi.feng/anaconda2/lib/python2.7/site-packages/chainer/datasets/sub_dataset.py", line 75, in get_example
return self._dataset[index]
File "/home/users/yuxi.feng/anaconda2/lib/python2.7/site-packages/chainer/dataset/dataset_mixin.py", line 30, in getitem
return self.get_example(index)
File "/home/users/yuxi.feng/Project/hyperface/scripts/datasets.py", line 293, in get_example
raise IndexError
IndexError
I used aflw datasets as your example mentined, it seems some exapmles has no proposal rects??

Hi takiyu,

I tried to use trained model (model_epoch_190 downloaded from your link). It seems there is an error when I use python ./scripts/use_on_file.py --model model_epoch_190 --img sample_images/lena_face.png

I have looked into model_epoch_190. There are all .npy files in that folder. But it needs chainer.serializers.load_npz(args.model, model).

Am I using the trained model in a right way? Could you tell me more details about how I can use model_epoch_190, please?

hi, thank you for sharing the code.
Could you share me with aflw.sqlite and trained model ?Because I cannot download it from the link.
my email is [email protected]
thank you very much

I want to test my image with pretrained model. After download your trained model and unzip, I try "python ./scripts/use_on_file.py --model model_epoch_190 --img sample_images/my_image.png".
But i get some error like below.
File "./scripts/use_on_file.py", line 57, in
chainer.serializers.load_npz(args.model, model)
File "/usr/local/lib/python2.7/dist-packages/chainer/serializers/npz.py", line 150, in load_npz
with numpy.load(file) as f:
File "/home/sungkwonan/.local/lib/python2.7/site-packages/numpy/lib/npyio.py", line 372, in load
fid = open(file, "rb")
IOError: [Errno 21] Is a directory: 'model_epoch_190'

how can i set the trained model?

Hi,

Sometimes, when I train a model , I get the following error:

WARNING (29033) datasets : Invalid crop rectangle. (rect:(1209, 0, 281, 169), img:(170, 0, 3), org_img(1600, 1142, 3)")

Do you know what could be the cause ?

I don't know what's the range of roll, pitch and yaw, so I couldn't convert the float value to specific angle? Does all of them is between [-90, 90] ? Or[-100,100] for yaw, [-45,45] for pitch, [-25, 25] for roll ? Or some other function I don't know ?
And why the Yaw value reading from db is not between [-1., 1.] ? For example:
Roll: 0.21319378912448883
Pitch: 0.0004151661123614758
Yaw: -1.1322304010391235
x: 198
y: 55
w: 220
h: 220

Hello takiyu，

Thank you for the code you provided, but I cannot find the file "aflw_cache.npz", how can I get this file.
Thanks!

Hi, many thanks for the good work! (will you implement the latest work 'all in one cnn for face analysis' as well?)
Anyway, by running use_on_cap or demo_on_test or use_on_file etc the output (landmarks, rectangle ..) is only for ONE face although there are more then one clear and focused faces in the image.
How can I change it?

Hi takiyu,
I am trying to run your implementation of Hyperface but I can't see the file "aflw_cache.npz".
Thanks.
Cheers,
Anna

Hi, @takiyu
I'm running the use_on_file.py like this
python scripts/use_on_file.py --model model_epoch_190 --img age_gender/input/jan2.jpg

The error I'm getting is this:

Traceback (most recent call last):
  File "scripts/use_on_file.py", line 87, in <module>
    y = model(x)
  File "/datadrive/hyperface/scripts/models.py", line 80, in __call__
    h = F.dropout(h, train=self.train)
  File "/datadrive/virtualenvs/py2/local/lib/python2.7/site-packages/chainer/functions/noise/dropout.py", line 163, in dropout
    kwargs, train='train argument is not supported anymore. '
  File "/datadrive/virtualenvs/py2/local/lib/python2.7/site-packages/chainer/utils/argument.py", line 7, in check_unexpected_kwargs
    raise ValueError(message)
ValueError: train argument is not supported anymore. Use chainer.using_config

The model is successfully running till this step:
[2018/10/12 04:45:15] INFO (4127) __main__ : Forward the network

Is there some step I'm missing out?