huangziliandy / rpnsd Goto Github PK

I tried to train the model on the AMI corpus, but the loss didn't seem to converge

Thanks for this great job and for making it public. I'm new in this domain and I'm trying to test my own data using your trained model. The steps are as follow:
1- Installing Kaldi and Faster-rcnn.
2- Downloading the modelbest.pth.tar file under RPNSD/model
3- Running ./inference.sh

The output that I got :
Experiment directory is experiment/pretraincfgres101epoch1bs8opsgdlr0.01min_lr0.0001schedulermultipat10seed7alpha1.0archres101dev12000modelbestfreeze0bnfix0cfgres101epoch10bs8opsgdlr0.00004min_lr0.00004pat10seed7alpha0.1archres101
Decision threshold is 0.5
NMS threshold is 0.3
Fold 1 Modelname is modelbest
scripts/eval_cpu.sh: line 17: parse_options.sh: No such file or directory

Am I missing something?
Should I adapt my data first then run the inference?

First of all thanks for RPNSD. but the format of the file provided by you is not extracting anything says invalid compressed file.
Can you reupload it???

Hello！
I try to download the pretrained model (https://drive.google.com/file/d/1EYhTADveeeMlu2J3AqzkITcKXZhbNmUa/view). But i cannont open the [modelbest.pth.tar] by WinRar. And it shows that the file is destoryed. How can i download the pretrained model by other methods ？

Has anyone sucessfully obtained the RPNSD pretrained model that shared on this Git repo? For me, the downloaded archive's gotten damaged and failed to be decompressed.

Specifically what files should I have and what should be their format, so I can run prepare_callhome_5folds.sh from stage 1 on my own dataset?

Hi,

Where can I find the ROI Pooling code?
There are these lines In faster_rcnn.py↓↓
from model.roi_pooling.modules.roi_pool import _RoIPooling
from model.roi_crop.modules.roi_crop import _RoICrop
from model.roi_align.modules.roi_align import RoIAlignAvg

Could you please share the code for roi_pooling to fix this issue? Or am I missing something?

RoI Poolingのコードが見当たりませんが、なくても問題なく動作するのでしょうか？
以下のラインによるとmodel以下のディレクトリにroi_~~というディレクトリが存在すると思うのですが、見当たりません。
from model.roi_pooling.modules.roi_pool import _RoIPooling
from model.roi_crop.modules.roi_crop import _RoICrop
from model.roi_align.modules.roi_align import RoIAlignAvg

コードを見せていただくことはできませんでしょうか？
（英語がへたくそなので念のため日本語で。よろしくお願いいたします。）

Hello, thank you for sharing your research results.
We have adapted to Callhome DB with the pre-trained model you shared.
This is 2% higher than the DER value reported in the paper.

Could you please check it?

I am trying to train a model from scratch but I realized that "from model.nms.nms_wrapper import nms" lines (e.g. in model/rpn/proposal_layer.py) fail as the repo does not include nms directory under scripts/model/. As far as I understand, cluster_nms.py is also not the correct one as it also calls model.nms. Could you please share the code for NMS to fix this issue? Or am I missing something?

Hi, I am interested in your solution, I would appreciate if you have time to answer to some of my questions:
I have notices there were some lacking files which I complete from jwyang's implementation. Afterwards I procedeed with the training and I keep obtaining the same error:

ValueError: bg_num_rois = 0 and fg_num_rois = 0, this should not happen!

Any idea why is that happening?

When using the adaptation script using my own dataset I get this error:

return torch._C._nn.log_softmax(input, dim)
RuntimeError: dimension out of range (expected to be in range of [-1, 0], but got 1)

• it's better to acknowledge the faster rcnn repo you used (https://github.com/jwyang/faster-rcnn.pytorch.git) in README.md
• add the license information
• please take care of CLSP specific parts (e.g., free-gpu -n 1, https://github.com/HuangZiliAndy/RPNSD/blob/master/path.sh#L7-L9, https://github.com/HuangZiliAndy/RPNSD/blob/master/scripts/swbd_sre/prepare_swbd_sre.sh#L13)

Hi,

Thanks for making RPNSD available. I was wondering, what if I don't want to run the experiment? Say I have a long audio file, file.wav, can I just run RPNSD on it and get the diarization result? For example, something like:

  ./diarize.py  --pretrained_model modelbest.pth.tar file.wav

Thanks