# CrossModel

**Repository Path**: xeonye/CrossModel

## Basic Information

- **Project Name**: CrossModel
- **Description**: No description available
- **Primary Language**: Unknown
- **License**: Not specified
- **Default Branch**: images
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 0
- **Created**: 2024-12-10
- **Last Updated**: 2024-12-10

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

# HARNN
Mentioned of the part of **HARNN** model, we use the implementation of https://github.com/electron1c/HARNN-pytorch.

Details can be seen by visiting the previous link.

# CN-Clip

As for the CN-Clip part, we use the library provided by https://github.com/OFA-Sys/Chinese-CLIP. And install it using pip.

We only use the CN-Clip model to retrieve feature vectors of texts and images to feed to HARNN.

Some changes have been made to CN-Clip, as follows.

First find the method CLIP.encode_text(), and make changes to it in order to return the text features(not the global features).

![img.png](img.png)