search

junyanz / pytorch-CycleGAN-and-pix2pix

Image-to-Image Translation in PyTorch
Other
23.09k stars 6.32k forks source link

How to retrain model using saved weights? #840

Closed MuhammadAsadJaved closed 5 years ago

MuhammadAsadJaved commented 5 years ago

Hi , I have trained a pix2pix model on custom dataset for 200 epochs and tested it but the generated images are not very good. can you give me some suggestion to increase the performance? 1-I have trained model on 512 x 512 visual images to generate infrared images using paired images. 2- How can I continue the training from last stored weights? Do you think the performance will increase if i train this model more ? i.e for 1000 epoch. 3-Please see few attached input and output examples generated with trained model. 3- Any other suggestions to increase the quality of output? 1 1output 2 2output 3 3output 4 4output

junyanz commented 5 years ago

See this Q & A regarding continue training. For better results, you can try our recent work such as pix2pixHD and SPADE.

MuhammadAsadJaved commented 5 years ago

Thank you very much.