---
license: apache-2.0
language:
- en
base_model:
- timm/ViT-SO400M-14-SigLIP-384
library_name: open_clip
tags:
- genshin-impact
- game
- clip
- siglip
- hoyoverse
---
# GenshinCLIP
A simple open-sourced [SigLIP model](https://huggingface.co/timm/ViT-SO400M-14-SigLIP-384) further fine-tuned on 15k Genshin Impact (Game Version 5.0) English text-image pairs at resolution 384x384.

Visit the [github](https://github.com/mrzjy/GenshinCLIP) for case study and data pair examples.

The model is **far from being perfect**, but could still offer some better text-image matching performance in some Genshin Impact scenarios.

- Deprecated Models (Trained on Genshin Impact Version 4.x)

| Model                                                                                     | Checkpoint Size | Val Loss |
|:-------------------------------------------------------------------------------------------:|:-----------------:|:----------:|
| [GenshinImpact-CLIP-ViT-B-16-laion2B-s34B-b88K](https://huggingface.co/mrzjy/GenshinImpact-CLIP-ViT-B-16-laion2B-s34B-b88K) | 0.59 GB         | 1.152    |
| [GenshinImpact-ViT-SO400M-14-SigLIP-384](https://huggingface.co/mrzjy/GenshinImpact-ViT-SO400M-14-SigLIP-384)        | 3.51 GB         | 0.362    |

## 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:

```python
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-5.0-ViT-SO400M-14-SigLIP-384')
tokenizer = get_tokenizer('hf-hub:mrzjy/GenshinImpact-5.0-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)
```

## Model Card
### SigLIP for GenshinImpact

[SigLIP model](https://huggingface.co/timm/ViT-SO400M-14-SigLIP-384) further fine-tuned on 15k Genshin Impact English text-image pairs at resolution 384x384.

(Note: A slight error occurred in counting training samples for deprecated models. The latest training data should actually include 3,500 additional samples compared to previous counts.)

### Training data description

There're currently 14,573 (train) and 298 (validation) text-image pairs used for model training.

All the images and texts are crawled from [Genshin Fandom Wiki](https://genshin-impact.fandom.com/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.

**Data Distribution:**

![data_distribution.png](img%2Fdata_distribution.png)