search

cristinalunaj / MMEmotionRecognition

Repository with the code of the paper: A proposal for Multimodal Emotion Recognition using auraltransformers and Action Units on RAVDESS dataset
MIT License
96 stars 37 forks source link

Incoherence size indicated and size provided in forward of Sequential Model #5

Closed nathanb97 closed 2 years ago

nathanb97 commented 2 years ago

Hello, First of all, thank you for your excellent paper, very pleasant to read, and your code too. It's quite rare to have a clear enough code with a good paper

However there are some small errors in your code that prevent me from trying yours models.

For the Facial Emotion Recognition For the sequential model for Video/models/sequenceLearning/workflow/run.py Following the MMEmotionRecognition/src/Video/models/sequenceLearning/README_AUS.md

The error that prevents me from continuing to continue is the following: 

    raise RuntimeError(
RuntimeError: input.size(-1) must be equal to input_size. Expected 35, got 714

In forward of LSTMEncoder()

Indeed a csv has 714 columns, did you select all of them ?

Because in your config file You had indicated 35 in input_size. Thank you for your answer.

cristinalunaj commented 2 years ago

Hello, thanks for your comments :) I think that the error is because the model is expecting 35 columns, that should be the 35 AUs extracted from OpenFace (all the AUXX_c and AUXX_r) after running the python file AUsFeatureExtractor.py, i didn't include it in the tutorial of the sequenceLearning, but it is in the main README under Facial Emotion Recognition. Sorry for the inconvenience.

nathanb97 commented 2 years ago

Hello Cristina, No problem. I missed it. Thank you very much for the reply. I close the issue.