Hi! I am confused about the training of multi-size images. Based on the paper,
In other words, during training we implement the varying-input-size SPP-net by two fixed-size networks that share parameters
we train each full epoch on one network, and then switch to the other one (keeping all weights) for the next full epoch
My interpretion is as below: Provided the epoch number is 15 and the size of training images is classified 3 categories, like 180, 224, 250. In the training process, epoch 0 is trained using images with the size of 180, epoch 1 is trained using images with the size of 224, and epoch 2 is trained with the size of 250. That is, the model is trained using the images with the size of 180, 224, 250 in order, right?
Hi! I am confused about the training of multi-size images. Based on the paper,
My interpretion is as below: Provided the epoch number is 15 and the size of training images is classified 3 categories, like 180, 224, 250. In the training process, epoch 0 is trained using images with the size of 180, epoch 1 is trained using images with the size of 224, and epoch 2 is trained with the size of 250. That is, the model is trained using the images with the size of 180, 224, 250 in order, right?
Thanks in advance! Looking forward to your help.