I have a question regarding the training for PackNetSAN. As stated in you paper, the training schedule includes 30 epochs where only the depth prediction network is trained, followed by 20 epochs where the depth prediction encoder is frozen and only the depth completion encoder and shared decoder are trained. I'm wondering any parameters can I set in the config file to make it happen in one training schedule, or it's actually done in two training schedules. Thank you!
Thank you for the great work!
I have a question regarding the training for PackNetSAN. As stated in you paper, the training schedule includes 30 epochs where only the depth prediction network is trained, followed by 20 epochs where the depth prediction encoder is frozen and only the depth completion encoder and shared decoder are trained. I'm wondering any parameters can I set in the config file to make it happen in one training schedule, or it's actually done in two training schedules. Thank you!