training error (seg_tri)

[leaf918]

loss graph didn't close.

[nanthakumar94]

[2018/12/13 09:22:03.066] INFO (15628) config : Load config from "config.json"
[2018/12/13 09:22:03.125] WARNING (15628) datasets.common : no image is valid
[2018/12/13 09:22:03.126] INFO (15628) main : Compute weight matrix for each image

i got this one.i don't know where is it error. please tell me

[leaf918]

[2018/12/13 09:22:03.066] INFO (15628) config : Load config from "config.json"
[2018/12/13 09:22:03.125] WARNING (15628) datasets.common : no image is valid
[2018/12/13 09:22:03.126] INFO (15628) main : Compute weight matrix for each image

i got this one.i don't know where is it error. please tell me

you need prepare all dataset first.

[kkkmax]

[2018/12/13 09:22:03.066] INFO (15628) config : Load config from "config.json"
[2018/12/13 09:22:03.125] WARNING (15628) datasets.common : no image is valid
[2018/12/13 09:22:03.126] INFO (15628) main : Compute weight matrix for each image
i got this one.i don't know where is it error. please tell me

you need prepare all dataset first.

Brother, my three modes of training:
$ python scripts/train.py --mode seg
$ python scripts/train.py --mode seg+
$ python scripts/train.py --mode seg_tri

The training part looks normal, but the test was very bad and found:
Mode=seg
Mode=seg+
The test result is a gray picture

Mode = seg_tri
The test result is a white picture

My test command is:
$ python scripts/train.py --mode {seg,seg+,seg_tri} --model_path './result/model_best' -i

Can you see where I am using the error? such as:

1. Is the model path I loaded correct?
2. Like you said W B needs to be changed to standardized initialization
3. Testing or training There is a particularly critical detail to note?

If you see a reply, I will be very grateful and happy!
@leaf918

[kkkmax]

@leaf918

In addition, mode=seg_tri training model, my loss curve is almost the same as you. Do you have some new opinions on this now?

[mansentan]

[2018/12/13 09:22:03.066] INFO (15628) config : Load config from "config.json"
[2018/12/13 09:22:03.125] WARNING (15628) datasets.common : no image is valid
[2018/12/13 09:22:03.126] INFO (15628) main : Compute weight matrix for each image
i got this one.i don't know where is it error. please tell me

you need prepare all dataset first.

@leaf918, do you know how to generate "img_alpha"?

[leaf918]

[2018/12/13 09:22:03.066] INFO (15628) config : Load config from "config.json"
[2018/12/13 09:22:03.125] WARNING (15628) datasets.common : no image is valid
[2018/12/13 09:22:03.126] INFO (15628) main : Compute weight matrix for each image
i got this one.i don't know where is it error. please tell me

you need prepare all dataset first.

@leaf918, do you know how to generate "img_alpha"?

python scripts/prepare_matting_dataset.py --pseudo_alpha

[leaf918]

@leaf918

In addition, mode=seg_tri training model, my loss curve is almost the same as you. Do you have some new opinions on this now?

not yet

[billzw]

alembic==1.0.10
asn1crypto==0.24.0
blurhash==1.1.0
boost==0.1
certifi==2019.3.9
cffi==1.12.3
chainer==5.4.0
chainerui==0.9.0
chardet==3.0.4
Click==7.0
cmake==3.13.3
colorama==0.4.1
cryptography==2.6.1
cupy-cuda100==6.0.0
cycler==0.10.0
decorator==4.4.0
dlib==19.9.0
fastrlock==0.4
filelock==3.0.10
Flask==1.0.3
gevent==1.4.0
greenlet==0.4.15
h5py==2.9.0
http-ece==1.1.0
idna==2.8
itsdangerous==1.1.0
Jinja2==2.10.1
kiwisolver==1.1.0
logutils==0.3.5
Mako==1.0.10
MarkupSafe==1.1.1
Mastodon.py==1.4.2
matplotlib==3.1.0
msgpack==0.6.1
numpy==1.16.2
opencv-python==3.4.5.20
Pillow==6.0.0
protobuf==3.7.1
pycparser==2.19
pyparsing==2.4.0
python-dateutil==2.8.0
python-editor==1.0.4
python-magic==0.4.15
pytz==2019.1
rainbow-logging-handler==2.2.2
requests==2.22.0
scipy==1.0.0
six==1.12.0
SQLAlchemy==1.3.3
structlog==19.1.0
typing==3.6.6
typing-extensions==3.7.2
urllib3==1.25.2
Werkzeug==0.15.4

[leaf918]

could you paste some result image here , @billzw

[billzw]

[Maru-chan511]

@billzw hi~does the last black picture is your test output?

[leaf918]

@billzw hi~does the last black picture is your test output?

yes.

[Maru-chan511]

@leaf918 you have two github ID? Or you got the same output?