Baseline 4

[Ad7amstein]

• Implementation 1
  • Use the classifier of B1-tuned to extract representation per clip
  • Train an LSTM on the clip (9 steps) sequences
• Implementation 2
  • Extend the classifier network directly with an LSTM layer then classification