Calculate width (max of all) and height (sum) for a vertical merge of images. Args: images (List[PIL.Image or np.ndarray]): List of

Args: logits: [minibatch, C, T], not passed to the softmax func. targets, torch.cuda.LongTensor [minibatch, T]

Calculate the IoU of lines between two bounding boxes. Args: bbox1 (List[float]): First bounding box [x_min, y_min, x_max, y_max].

Check if layout model exists, download if missing. Args: model_path: Path to layout model file auto_download: If True, automatica

Check if model files exist, download if missing. Args: encoder_path: Path to encoder ONNX model decoder_path: Path to decoder ONN

Check if the input parameters are valid based on the initialized models. Args: input_params (Dict): A dictionary contain

3x3 convolution with padding.

3x3 convolution with padding.

Convert applicable model parameters to fp16

Create LMDB dataset for training and evaluation. ARGS: inputPath : input folder path where starts imagePath outputPath : LMD

Create Gradio demo interface. Returns: gr.Blocks: Gradio Blocks application instance.

Create openocr package structure and copy necessary files

CTC decoder using multiple processes.

CTC greedy (best path) decoder.

convert text-index into text-label.

Single-threaded text detection for all images.

Drop paths (Stochastic Depth) per sample (when applied in main path of residual blocks). This is the same as the DropConnect impl I created f

convert text-label into text-index.

convert text-label into text-index. input: text: text labels of each image. [batch_size] output: text: concat

convert text-label into text-index. input: text: text labels of each image. [batch_size] output: text: concat

Encodes the given text into a list of character IDs and generates various lists for question and prompt sequences. Args:

convert text-label into text-index. input: text: text labels of each image. [batch_size] output: text: concat

convert text-label into text-index. input: text: text labels of each image. [batch_size] output: text: concat

expand poly along width.

expand poly along width.

Export TableData to HTML table. Args: table_data (TableData): TableData object. Returns: str: HTML string.

f_direction: h x w x 2 pos_list: [[y, x], [y, x], [y, x] ...]

Extract and clean table tags from HTML string

Detect if string ends with a repeating phrase. Args: s (str): Input string. min_len (int): Minimum length of unit. m

Find the shortest substring that repeats to form the entire string. Args: s (str): Input string. Returns: str or None:

flag.

Format markdown text for display. This function handles: - HTML tables (pass through as-is for Gradio Markdown rendering) - LaTeX formula

Forward pass for a single stream (i.e. content or query) tgt_norm is just a LayerNorm'd tgt. Added as a separate parameter for efficiency.

Translation work in one batch.

Cloze Mask Decoding.

Perform the forward pass for the test phase. Args: x (torch.Tensor): Input tensor of shape (batch_size, channels, height