readme_content = """# Image Captioning Model
This is an Image Captioning model trained on a custom dataset with 6000 training images and 1000 validation images. It uses ResNet50 for image feature extraction and an LSTM with fine-tuned GloVe embeddings for caption generation.
Model Details
- Image Encoder: ResNet50
- Text Decoder: LSTM with fine-tuned GloVe embeddings
- Vocabulary Size: 3478
- Embedding Dimension: 512
- Hidden Size: 1024
- Number of Layers: 2
- Max Caption Length: 30
Usage
Load the model and tokenizer using PyTorch and the provided model.py.
Example
import torch
from PIL import Image
import pickle
from huggingface_hub import hf_hub_download
from model import image_captioning_model
# Download files
repo_id = "Pumponchik/image_caption_weak"
model_path = hf_hub_download(repo_id=repo_id, filename="pytorch_model.bin")
config_path = hf_hub_download(repo_id=repo_id, filename="config.json")
tokenizer_path = hf_hub_download(repo_id=repo_id, filename="tokenizer.pkl")
transform_path = hf_hub_download(repo_id=repo_id, filename="image_transforms.pkl")
# Load config
with open(config_path, "r") as f:
config = json.load(f)
# Load model
model = image_captioning_model(**config)
model.load_state_dict(torch.load(model_path))
model.eval()
device = "cuda" if torch.cuda.is_available() else "cpu"
model.to(device)
# Load tokenizer and transforms
with open(tokenizer_path, "rb") as f:
tokenizer = pickle.load(f)
with open(transform_path, "rb") as f:
transform = pickle.load(f)
# Generate caption
image = Image.open("example.png").convert("RGB")
image_tensor = transform(image).unsqueeze(0).to(device)
caption = model.generate_caption(image_tensor, tokenizer, max_length=config["max_len"], device=device)
print("Generated Caption:", caption)