Task1 uses image embedding to predict the pathologies. Task2 uses images embedding and report to predict the pathologies
The 'task1_v2_cache.py' is the final version our group uses for Task 1.
The 'task1_with_AutoEncoder_with_embeding_use_cache.py' explores combining information from the 'mimic-cxr-2.0.0-metadata' with image embeddings as training input. However, the model's performance was suboptimal.