Multimodel Fusion Strategies: LSTM, BERT, CNN, ResNet

This repository explores multimodal fusion strategies using a combination of LSTM, BERT, CNN, and ResNet models for improved performance in classification tasks.

Fusion Strategies

We implement and compare two types of fusion strategies:

🔹 Late Fusion (Decision-Level Fusion)

In this approach, outputs from different models are fused at the decision level (e.g., after classification probabilities are generated).

1. LSTM + CNN
2. LSTM + ResNet
3. BERT + CNN
4. BERT + ResNet

🔹 Early Fusion (Feature-Level Fusion)

Here, features from different models are combined before classification.

5. LSTM + CNN
6. LSTM + ResNet
7. BERT + CNN
8. BERT + ResNet