Closed martinwholtmon closed 2 weeks ago
Implement the last model using a pre-trained CNN to extract features (spatial information), and then use an RNN to capture temporal information.
CNN: BNInception or inception v3 (Pre-trained on K100 RGB): https://yjxiong.me/others/kinetics_action/
RNN: LSTM probably, look into it.
Implement the last model using a pre-trained CNN to extract features (spatial information), and then use an RNN to capture temporal information.
CNN: BNInception or inception v3 (Pre-trained on K100 RGB): https://yjxiong.me/others/kinetics_action/
RNN: LSTM probably, look into it.