Datasets
The following table summarizes the supported tasks, datasets, their respective modalities, and indicates whether the dataset is automatically downloaded by the system (marked as “Auto”):
Task |
Dataset |
Modality |
Auto |
|---|---|---|---|
Object Classification |
CIFAR10 |
Image |
Yes |
Object Classification |
TinyImageNet |
Image |
No |
Traffic Sign Recognition |
GTSRB |
Image |
Yes |
Facial Recognition |
CelebA |
Image |
Yes |
Sentiment Analysis |
SST-2 |
Text |
Yes |
Sentiment Analysis |
IMDb |
Text |
Yes |
Topic Classification |
DBpedia |
Text |
Yes |
Topic Classification |
AG’s News |
Text |
Yes |
Speech Command Recognition |
SpeechCommands |
Audio |
Yes |
Music Genre Classification |
GTZAN |
Audio |
Yes |
Speaker Identification |
VoxCeleb1 |
Audio |
Yes |
Video Classification |
HMDB51 |
Video |
No |
Audiovisual Sentiment Analysis |
HMDB51 |
Audiovisual |
No |
Visual Question Answering (VQA) |
HMDB51 |
Image/Text |
No |
Contrastive Learning |
CIFAR10 + STL-10 |
Image/Contrastive |
No |