Skip to content

Request for Recommended Hyperparameter Settings for Segmentation Datasets #1

@MYMY-young

Description

@MYMY-young

Thank you for your valuable research and contributions.

Currently, the README provides an example command as follows:

python generate.py --seed=42 --domain_emb_loc="outputs/emb/doge.pt" --gen_src_dir="data/domainnet/real/angel" --save_dir="outputs/domainnet/angel" --edit_weight=1.5 --n_batch=10 --n_per_prompt=1 --num_inference_steps=20 --save_grid --device="cuda:0" --neg_prompt "blurry, blurred, ambiguous, blending, opaque, translucent, layering, shading, mixing, ugly, tiling, poorly drawn face, out of frame, mutation, disfigured, deformed, blurry, bad art, bad anatomy, text, watermark, grainy, underexposed, unreal architecture, unreal sky, weird colors" --guidance_scale=5.0 --control_type="canny" --controlnet_conditioning_scale=1.0 --canny_lo_thres=100 --canny_hi_thres=800

Could you please provide recommended hyperparameter settings that would be suitable for segmentation tasks?

Thanks.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions