Image Search-TPU

ImageSearch-TPU is a text-based image search application that uses the Image CLIP model, ported to SG2300X chip series products using the Sophon SDK for local TPU hardware-accelerated inference. It enables quick searching of image libraries based on text using a large number of existing images and provides a user-friendly interaction interface using Streamlit.

Clone the repository:

git clone https://github.com/zifeng-radxa/ImageSearch-tpu

Download the CLIP bmodel:

cd ImageSearch-tpu/clip_image_search/clip
wget https://github.com/radxa-edge/TPU-Edge-AI/releases/download/ImageSearch/ImageSearch_bmodel.zip
unzip ImageSearch_bmodel.zip

Place the bmodel in ImageSearch_bmodel/bmodels/.

rm -rf bmodels/
mv ./ImageSearch_bmodel/bmodels/ .

The file tree structure will be as follows:

.
└── ImageSearch-tpu
    ├── __pycache__
    ├── clip_image_search
    │   ├── __pycache__
    │   └── clip
    │       ├── ImageSearch_bmodel
    │       ├── __pycache__
    │       ├── bmodels
    │       │   ├── CH
    │       │   └── EN
    │       └── saved_tokenizer
    │           ├── bert_chinese_tokenizer-fast
    │           └── bert_chinese_tokenizer-slow
    ├── gallery_collection
    ├── images
    └── results
        ├── CH
        └── EN

Create a virtual environment:

It's necessary to create a virtual environment; otherwise, it may affect the normal operation of other applications. For virtual environment usage, please refer here.
```
cd ImageSearch-tpu
python3 -m virtualenv .venv
source .venv/bin/activate
```

Install dependencies:

pip3 install --upgrade pip
pip3 install -r requirements.txt
pip3 install https://github.com/radxa-edge/TPU-Edge-AI/releases/download/v0.1.0/tpu_perf-1.2.31-py3-none-manylinux2014_aarch64.whl

Start the web service:

streamlit run app.py EN # for CLIP-VIT-B32
streamlit run app.py CH # for ChineseCLIP VIT-B16

Access port 8501 of the Airbox IP address in a web browser.

Image Search-TPU

Application Display​

Application Display