Imscore

Work by RE-N-Y and friends @ krea.ai

Imscore is a minimal library curating a set of fully differentiable aesthetic and preference scorers for images. We provide a set of popular scorers such as PickScore, MPS, HPSv2, and LAION aesthetic scorer as well as our own models trained on open source preference datasets.

imscore allows ...

1. Benchmarking your generative models on aesthetic and preference scorers.
2. Post training your generative models to align with human preference.
3. Prevent headaches of porting over models from different repositories.

Installation

pip install imscore

Usage

from imscore.aesthetic.model import ShadowAesthetic, LAIONAestheticScorer
from imscore.hps.model import HPSv2
from imscore.mps.model import MPS
from imscore.preference.model import SiglipPreferenceScorer
from imscore.pickscore.model import PickScorer

import torch
import numpy as np
from PIL import Image
from einops import rearrange

# popular aesthetic/preference scorers
model = ShadowAesthetic() # ShadowAesthetic aesthetic scorer (my favorite)
model = PickScorer() # PickScore preference scorer
model = MPS.from_pretrained("RE-N-Y/mpsv1") # MPS (ovreall) preference scorer
model = HPSv2.from_pretrained("RE-N-Y/hpsv21") # HPSv2.1 preference scorer
model = LAIONAestheticScorer.from_pretrained("RE-N-Y/laion-aesthetic") # LAION aesthetic scorer

# multimodal (pixels + text) preference scorers trained on PickaPicv2 dataset 
model = SiglipPreferenceScorer.from_pretrained("RE-N-Y/pickscore-siglip")

prompts = "a photo of a cat"
pixels = Image.open("cat.jpg")
pixels = np.array(pixels)
pixels = rearrange(torch.tensor(pixels), "h w c -> 1 c h w") / 255.0

# prompts and pixels should have the same batch dimension
# pixels should be in the range [0, 1]
# score == logits
score = model.score(pixels, prompts) # full differentiable reward

Post Training for Generative Models

import torch
from imscore.preference.model import SiglipPreferenceScorer

G = model() # your generative model
dataloader = ... # your dataloader with conditioning (ex.prompts)

rm = SiglipPreferenceScorer.from_pretrained("RE-N-Y/pickscore-siglip") # pretrained preference model
optim = torch.optim.AdamW(G.parameters(), lr=3e-4)

# post training
for prompts in dataloader:
    optim.zero_grad()
    
    images = G(prompts)
    scores = rm.score(images, prompts) # ensure images are in the range [0, 1]
    loss = -scores.mean() # maximise reward

    loss.backward()
    optim.step()

List of available models

Aesthetic Scorers

from imscore.aesthetic.model import ShadowAesthetic, CLIPAestheticScorer, SiglipAestheticScorer, Dinov2AestheticScorer, LAIONAestheticScorer

# pixel only scorers trained on imscore dataset's aesthetic rating
SiglipAestheticScorer.from_pretrained("RE-N-Y/imreward-fidelity_rating-siglip")
CLIPAestheticScorer.from_pretrained("RE-N-Y/imreward-fidelity_rating-clip")
Dinov2AestheticScorer.from_pretrained("RE-N-Y/imreward-fidelity_rating-dinov2")

# pixel only scorers trained on imreward dataset's overall rating
SiglipAestheticScorer.from_pretrained("RE-N-Y/imreward-overall_rating-siglip")
CLIPAestheticScorer.from_pretrained("RE-N-Y/imreward-overall_rating-clip")
Dinov2AestheticScorer.from_pretrained("RE-N-Y/imreward-overall_rating-dinov2")

# pixel only scorers trained on AVA dataset
CLIPAestheticScorer.from_pretrained("RE-N-Y/ava-rating-clip-sampled-True")
CLIPAestheticScorer.from_pretrained("RE-N-Y/ava-rating-clip-sampled-False")
SiglipAestheticScorer.from_pretrained("RE-N-Y/ava-rating-siglip-sampled-True")
SiglipAestheticScorer.from_pretrained("RE-N-Y/ava-rating-siglip-sampled-False")
Dinov2AestheticScorer.from_pretrained("RE-N-Y/ava-rating-dinov2-sampled-True")
Dinov2AestheticScorer.from_pretrained("RE-N-Y/ava-rating-dinov2-sampled-False")

# Common aesthetic scorers
LAIONAestheticScorer.from_pretrained("RE-N-Y/laion-aesthetic") # LAION aesthetic scorer
ShadowAesthetic() # ShadowAesthetic aesthetic scorer for anime images

Preference Scorers

from imscore.hps.model import HPSv2
from imscore.mps.model import MPS
from imscore.preference.model import SiglipPreferenceScorer, CLIPPreferenceScorer
from imscore.pickscore.model import PickScorer

HPSv2.from_pretrained("RE-N-Y/hpsv21") # HPSv2.1 preference scorer
MPS.from_pretrained("RE-N-Y/mpsv1") # MPS (ovreall) preference scorer
PickScorer() # PickScore preference scorer

# multimodal scorers trained on PickAPicv2 dataset
SiglipPreferenceScorer.from_pretrained("RE-N-Y/pickscore-siglip")
CLIPPreferenceScorer.from_pretrained("RE-N-Y/pickscore-clip")

Benchmarks

Dataset	Model	Performance
ImageReward (overall rating)	imreward-overall_rating-siglip	1.147
ImageReward (overall rating)	imreward-overall_rating-clip	1.213
ImageReward (overall rating)	imreward-overall_rating-dinov2	1.111
ImageReward (aesthetic rating)	imreward-fidelity_rating-siglip	0.857
ImageReward (aesthetic rating)	imreward-fidelity_rating-clip	0.859
ImageReward (aesthetic rating)	imreward-fidelity_rating-dinov2	0.845
AVA	ava-rating-clip-sampled-True	0.228
AVA	ava-rating-clip-sampled-False	0.236
AVA	ava-rating-siglip-sampled-True	0.292
AVA	ava-rating-siglip-sampled-False	0.276
AVA	ava-rating-dinov2-sampled-True	0.264
AVA	ava-rating-dinov2-sampled-False	0.265
LAION	laion-aesthetic	0.321
PickAPicv2 (test + no ties)	pickscore	69.67%
PickAPicv2 (test + no ties)	pickscore-siglip	67.68%
PickAPicv2 (test + no ties)	pickscore-clip	67.46%
???	ShadowAesthetic v2	???

Note : Imreward uses 1-7 likert scale, AVA uses 1-10 likert scale, PickAPicv2 asks model/user to pick the better image given a prompt. Shadow Aesthetic v2 is an anime aesthetic scorer with ViT backbone, training recipe and objective is not disclosed. For AVA, there's no official test set, so I've created my own split. The performance comparison between LAION aesthetic scorer and other scorers is not an apple-to-apple comparison.

Why did I make this?

1. To save myself headaches.
2. To provide a common interface for dataset filtering, posttraining, and image model benchmarking.

TODOs

• Benchmark scorers across ImageReward, PickAPicv2, MPS, and HPS datasets.
• Add AIMv2 backbone scorers.