am working on project , Facial Emotions recognition . I tried using two models RESNET50 and CNN from Github. its still confusing some emotions . am using FER2013 as my dataset can really use some help
Given that you’ve likely finished courses 2 and 3, please be specific in terms of what you need help with. Providing the following details will help someone with the time to help get a better idea of the issue:
- Your notebook
- Link to the model hosted on github.
- Details about the issue.
Kaggle seems to have this dataset with a lot of notebooks associated with it. Have you seen them?
am sorry for the confusion , i didn’t know i selected course 3 ;;
i completed .. Neural Networks and Deep Learning and Improving Deep Neural Networks: Hyperparameter Tuning, Regularization and Optimization
the kaggle dataset u linked is the dataset i have used
3.Its confusing emotions like sad and angry and neutral
4.i thought because of image imbalance in the dataset i thought its overfitting so i mixed two dataset affectnet and FER2013
sad 4599
disgust 2242
fear = 4452
anger 4815
happy 3419
neutral 4474
surprise 4290
these are the images per class
CUSTOM DATASET:
accuracy : 0.9317
acc 0.9591
f1 = 0.7364
loss = 0.6955
precision 0.8188
recall 0.6706
this is the result i got
but when testing emotion recognition was worse than just the FER2013 dataset