hackergeek/radiology-image-captioning

Radiology Image Captioning Model

Lightweight CNN+Transformer trained on eltorio/ROCO-radiology.

Usage

This is a custom PyTorch model. To load it, you would typically do the following:

from tokenizers import Tokenizer
import torch
# (Define your ImageCaptioningModel class as used during training)
# from my_model_definition import ImageCaptioningModel # if saved separately
tokenizer = Tokenizer.from_file('hackergeek/radiology-image-captioning/vocab.json')
config = json.load(open('hackergeek/radiology-image-captioning/config.json'))
# model = ImageCaptioningModel(vocab_size=config['vocab_size'], embed_dim=config['embed_dim'])
# model.load_state_dict(torch.load('hackergeek/radiology-image-captioning/pytorch_model.bin', map_location='cpu'))
# model.eval()