README.md

metadata

language: en
license: mit
library_name: transformers
tags:
  - image-to-text
widget:
  - src: >-
      https://huggingface.co/IAMJB/interpret-cxr-impression-baseline/resolve/main/effusions-bibasal.jpg
  - src: >-
      https://huggingface.co/IAMJB/interpret-cxr-impression-baseline/resolve/main/Chest-X-ray-taken-on-2-nd-day-of-admission-in-the_Q320.jpg
  - src: >-
      https://huggingface.co/IAMJB/interpret-cxr-impression-baseline/resolve/main/effusions-bibasal.jpg

Evaluation on chexpert-plus

Usage:

import torch
from PIL import Image
from transformers import BertTokenizer, ViTImageProcessor, VisionEncoderDecoderModel, GenerationConfig
import requests

mode = "findings"
# Model
model = VisionEncoderDecoderModel.from_pretrained(f"IAMJB/chexpert-mimic-cxr-{mode}-baseline").eval()
tokenizer = BertTokenizer.from_pretrained(f"IAMJB/chexpert-mimic-cxr-{mode}-baseline")
image_processor = ViTImageProcessor.from_pretrained(f"IAMJB/chexpert-mimic-cxr-{mode}-baseline")
#
# Dataset
generation_args = {
   "bos_token_id": model.config.bos_token_id,
   "eos_token_id": model.config.eos_token_id,
   "pad_token_id": model.config.pad_token_id,
   "num_return_sequences": 1,
   "max_length": 128,
   "use_cache": True,
   "beam_width": 2,
}
#
# Inference
refs = []
hyps = []
with torch.no_grad():
   url = "https://huggingface.co/IAMJB/interpret-cxr-impression-baseline/resolve/main/effusions-bibasal.jpg"
   image = Image.open(requests.get(url, stream=True).raw)
   pixel_values = image_processor(image, return_tensors="pt").pixel_values
   # Generate predictions
   generated_ids = model.generate(
       pixel_values,
       generation_config=GenerationConfig(
           **{**generation_args, "decoder_start_token_id": tokenizer.cls_token_id})
   )
   generated_texts = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)
   print(generated_texts)