license: apache-2.0
language:
- en
library_name: open_clip
pipeline_tag: zero-shot-image-classification
GenshinCLIP
A simple open-sourced SigLIP model fine-tuned on Genshin Impact's image-text pairs.
The model is far from being perfect, but could still offer some better text-image matching performance in some Genshin Impact scenarios.
Case Study
Case | Image | Multiple Choices | CLIPScore (After Finetune) |
CLIPScore (Before Finetune) |
---|---|---|---|---|
1 | ![]() |
1) This is an image of Kuki Shinobu 2) This is an image of Ganyu 3) This is an image of Keqing 4) This is an image of Liyue |
1) 0.000 2) 0.438 3) 0.000 4) 0.000 |
1) 1.207e-06 2) 4.144e-08 3) 1.201e-07 4) 2.212e-08 |
2 | ![]() |
1) This is an area of Liyue 2) This is an area of Mondstadt 3) This is an area of Sumeru 4) This is Qingce Village |
1) 0.016 2) 0.000 3) 0.001 4) 0.233 |
1) 0.015 2) 0.003 3) 0.009 4) 0.377 |
3 | ![]() |
1) This is Andrius Wolf of the North 2) This is Stormterror Dvalin 3) This is Guardian of Apep's Oasis |
1) 1.000 2) 0.000 3) 0.000 |
1) 0.001 2) 0.000 3) 0.000 |
4 | ![]() |
1) This is Amakumo Fruit 2) This is Padisarahs 3) This is Naku Weeds |
1) 0.000 2) 0.000 3) 0.024 |
1) 9.425e-07 2) 1.207e-06 3) 1.669e-05 |
Note: Case 4 is a bad case. (Correct answer is Amakumo Fruit)
Intended uses & limitations
You can use the raw model for tasks like zero-shot image classification and image-text retrieval.
How to use (With OpenCLIP)
Here is how to use this model to perform zero-shot image classification:
import torch
import torch.nn.functional as F
from PIL import Image
import requests
from open_clip import create_model_from_pretrained, get_tokenizer
def preprocess_text(string):
return "Genshin Impact\n" + string
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
# load checkpoint from local path
# model_path = "path/to/open_clip_pytorch_model.bin"
# model_name = "ViT-SO400M-14-SigLIP-384"
# model, preprocess = create_model_from_pretrained(model_name=model_name, pretrained=model_path, device=device)
# tokenizer = get_tokenizer(model_name)
# or load from hub
model, preprocess = create_model_from_pretrained('hf-hub:mrzjy/GenshinImpact-ViT-SO400M-14-SigLIP-384')
tokenizer = get_tokenizer('hf-hub:mrzjy/GenshinImpact-ViT-SO400M-14-SigLIP-384')
# image
image_url = "https://static.wikia.nocookie.net/gensin-impact/images/3/33/Qingce_Village.png"
image = Image.open(requests.get(image_url, stream=True).raw)
image = preprocess(image).unsqueeze(0).to(device)
# text choices
labels = [
"This is an area of Liyue",
"This is an area of Mondstadt",
"This is an area of Sumeru",
"This is Qingce Village"
]
labels = [preprocess_text(l) for l in labels]
text = tokenizer(labels, context_length=model.context_length).to(device)
with torch.autocast(device_type=device.type):
with torch.no_grad():
image_features = model.encode_image(image)
text_features = model.encode_text(text)
image_features = F.normalize(image_features, dim=-1)
image_features = F.normalize(image_features, dim=-1)
text_features = F.normalize(text_features, dim=-1)
text_probs = torch.sigmoid(image_features @ text_features.T * model.logit_scale.exp() + model.logit_bias)
scores = [f"{s:.3f}" for i, s in enumerate(text_probs.tolist()[0])]
print(scores) # [0.016, 0.000, 0.001, 0.233]
Model Card
SigLIP for GenshinImpact
SigLIP model further fine-tuned on 17k Genshin Impact English text-image pairs at resolution 384x384.
Training data description
There're currently 17,428 (train) and 918 (validation) text-image pairs used for model training.
All the images and texts are crawled from Genshin Fandom Wiki and are manually parsed to form text-image pairs.
Image Processing:
- Size: Resize all images to 384x384 pixels to match the original model training settings.
- Format: Accept images in PNG or GIF format. For GIFs, extract a random frame to create a static image for text-image pairs.
Text Processing:
- Source: Text can be from the simple caption attribute of an HTML
<img>
tag or specified web content. - Format: Prepend all texts with "Genshin Impact" along with some simple template to form natural language sentences.
For example, here are some training image-text pairs:
Image | Text |
---|---|
![]() |
Genshin Impact This is Tainted Water-Splitting Phantasm. Tainted Hydro Phantasm Idle Waving 2 |
![]() |
Genshin Impact This is Annihilation Specialist Mek. Although the Fontaine Research Institute of Kinetic Energy Engineering... ..., state apparatus enforcing the monopoly on violence. |
![]() |
Genshin Impact Collei Expressions a Smiley expression |
![]() |
Genshin Impact This is the map of viewpoint Nine Pillars of Peace in Cuijue Slope, Minlin, Liyue Nine shackles of stone were said to have been laid down deep in the valleys of Cuijue Slope to drive off evil and cleanse the world. |
Data Distribution:
Validation Loss Curve