PredRNN++ for Radar Return

Environment

• Python 3.6.15
• Pytorch 1.10.1

Data

Data path is /home/tintin/predrnn_model/predrnn_pp_radar_return/radar_image

• Type : Video sequences, each consisting of 6 frames. Each frame is a gray scale image(64*64 pixels) of radar return.
• Input : the first 4 frames of a sequence.
• Output : the 5th frame of a sequence.
• In/Output Shape : [batch_size,seq_lengh,channel,height,width]

Training

• Parameter
  • batch_size : 32
  • EPOCHS : 1000
  • lr : 0.001

Run main.py for training the model,

$ python main.py

After training, check ./result/ directory to get input and output images. To get detailed process in every epoch, run tensorboard.

tensorboard --logdir tensorboard-mse_radar --bind_all

The path for the trained model is /home/tintin/predrnn_model/predrnn_pp_radar_return/best_model_state_mse_radar.bin