yashk2810
Poor performance with time_inceptionV3_1.5987_loss.h5
When i try your weights name 'time_inceptionV3_1.5987_loss.h5', the performance of image-caption model is poor.
Taking the basketball image as a example, the output is as follows: """Normal Max search: man ."""
Have you met this problem before?
PS: My keras version is 2.1.2.
============================ I also load this weights on Keras 1.2.2 and Tensorflow 0.12.1, but it failed again. It seems the weights file 'time_inceptionV3_1.5987_loss.h5' is useless.
Yes, I am also facing this problem. You will have to retrain the network from scratch. I would suggest using Google Colab(if you don't have your own GPU) notebooks as they give free GPU compute time
@yashk2810 & @bigzhao I'm facing the same problem too. Are you getting better outputs after training from scratch again ?
Yes I'm getting good output with this weight file. Not upto the mark but yes I'm getting good output. May be you just try once to save as well as load the weight file(1.5987) after the last epoch of the training.
On Thu, 11 Oct 2018, 2:05 pm Murali Manohar Chowdary Kondragunta, < [email protected]> wrote:
@yashk2810 https://github.com/yashk2810 & @bigzhao https://github.com/bigzhao I'm facing the same problem too. Are you getting better outputs after training from scratch again ?
— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/yashk2810/Image-Captioning/issues/5#issuecomment-428869388, or mute the thread https://github.com/notifications/unsubscribe-auth/Ad0I8hGTJaTOXlxcE5KYM8O1MPfwJUhHks5ujwLsgaJpZM4S-L6j .
Hi there! Can you share the time_inceptionV3_2.8876_loss.h5 file? It seems that the DL link from the repository is outdated =(
