Deep Learning Emotion Classification Optimization -- 2

I'm working on an emotion classification project using Deep Learning, utilizing datasets such as TESS, RAVDESS, and EMODB. While I've achieved 100% accuracy with TESS using pretrained models like InceptionV3 and MobileNetV2, the other datasets have only yielded 30-40%.

I'm seeking assistance to improve my model's classification accuracy. This could involve refining the model architecture, tuning hyperparameters, or enhancing the training procedure.

The results from this project will need to be output as a confusion matrix.

Ideal skills and experience for this job include:
- Deep Learning
- Model optimization
- Experience with emotion classification datasets
- Proficiency in generating and interpreting classification report, loss function, and confusion matrices.