image-to-text: Salesforce/blip-image-captioning-base
image regression: ResNet in Pytorch
Clone this repo.
git clone https://github.com/TedYeh/CUPredictor.git
Run the main.py
to build the flask platform.
python main.py
Choose the photo you want to predict, and the model will give the result.
See detail at predictor.py
Model: ResNet50 (Pretrain=True)
Optimizer: AdamW (Learning Rate: 8e-4) 6 Epoch
Task: Binary Classification and Regression
Data: 2,058 photo stickers (see ./image dir)