search

machanic / AU_R-CNN

The official implementation code of paper: "AU R-CNN：Encoding Expert Prior Knowledge into R-CNN for Action Unit Detection".
https://arxiv.org/abs/1812.05788
87 stars 24 forks source link

How can I access your pre-trained model and image data? #7

Closed Indigo6 closed 4 years ago

Indigo6 commented 4 years ago

I tested your model using following command:

python train.py --eval_mode -bs 128 -e 1 --iteration 1

but encountered an error:

Traceback (most recent call last): File "train.py", line 460, in main() File "train.py", line 267, in main observation = evaluator.evaluate() File "/home/data3/fanglin/AU_R-CNN/AU_rcnn/extensions/AU_evaluator.py", line 50, in evaluate for idx, batch in enumerate(it): File "/usr/share/anaconda3/envs/enc/lib/python3.6/site-packages/chainer/iterators/serial_iterator.py", line 70, in next self._previous_epoch_detail = self.epoch_detail File "/usr/share/anaconda3/envs/enc/lib/python3.6/site-packages/chainer/iterators/serial_iterator.py", line 96, in epoch_detail return self.epoch + self.current_position / self._epoch_size ZeroDivisionError: division by zero

All dependencies are installed and all needed files are put in the right place

Indigo6 commented 4 years ago

I found this comes from no image data...

machanic commented 4 years ago

@Indigo6 Sorry for letting you wait , I have uploaded parts of my trained models to google drive , https://drive.google.com/open?id=1Q8tx98QIQnT1yz1uSrCxszkobs8MSgrY There remains some files have not been uploaded because the network is not good.

Indigo6 commented 4 years ago

@sharpstill Now I'm able to train AU_r-cnn on my dataset, but I come to wonder how I can compute my own 'mean_rgb.npy'? I find there are two datasets in your 'Training Necessary Files' with only one 'mean_rgb.npy', does that mean this file is AU-TASK-Universal?

machanic commented 4 years ago

@Indigo6 This mean of pixel is not very critical , you can try use my mean file. This is just a mean value over all images and all pixels.