Closed jS5t3r closed 1 year ago
https://github.com/sail-sg/MDT/blob/main/run_sample.sh includes the inference and evaluation. You can also modify https://github.com/sail-sg/MDT/blob/main/infer_mdt.py to inference images. You need a metric that defines the closeness between generated images and imagenet images, like FID or IS score. Then you can tune the cfg to make the metric as good as possible.
I want to generate images that are close to the imagenet dataset by using your provided weights:
Any suggestions how to do that?
Are these the scripts? https://github.com/sail-sg/MDT/blob/main/run_sample.sh