docparse / config.json
Kedhareswer's picture
docparse: OCR weights, code, benchmarks and model card
4f675c6 verified
Raw History Blame Contribute Delete
415 Bytes
{
"name": "docparse",
"framework": "pytorch",
"code": "docparse/",
"weights": "artifacts/docparse/ocr.pt",
"architecture": "LineOCR (CRNN: 6 conv layers + 2-layer BiLSTM, CTC)",
"parameters": 1484357,
"input": "grayscale text line, height 32, any width",
"charset": "printable ASCII + \u2022 \u20ac \u00a3 \u00b0 \u00a9",
"space": "https://huggingface.co/spaces/Kedhareswer/docparse"
}