# Copyright (C) 2021-2024, Mindee. # This program is licensed under the Apache License 2.0. # See LICENSE or go to for full license details. from copy import deepcopy from typing import Any, Dict, List, Optional, Tuple import tensorflow as tf from tensorflow.keras import Model, layers from doctr.datasets import VOCABS from doctr.models.classification import magc_resnet31 from doctr.models.modules.transformer import Decoder, PositionalEncoding from ...utils.tensorflow import _bf16_to_float32, load_pretrained_params from .base import _MASTER, _MASTERPostProcessor __all__ = ["MASTER", "master"] default_cfgs: Dict[str, Dict[str, Any]] = { "master": { "mean": (0.694, 0.695, 0.693), "std": (0.299, 0.296, 0.301), "input_shape": (32, 128, 3), "vocab": VOCABS["french"], "url": "https://doctr-static.mindee.com/models?id=v0.6.0/master-a8232e9f.zip&src=0", }, } class MASTER(_MASTER, Model): """Implements MASTER as described in paper: `_. Implementation based on the official TF implementation: `_. Args: ---- feature_extractor: the backbone serving as feature extractor vocab: vocabulary, (without EOS, SOS, PAD) d_model: d parameter for the transformer decoder dff: depth of the pointwise feed-forward layer num_heads: number of heads for the mutli-head attention module num_layers: number of decoder layers to stack max_length: maximum length of character sequence handled by the model dropout: dropout probability of the decoder input_shape: size of the image inputs exportable: onnx exportable returns only logits cfg: dictionary containing information about the model """ def __init__( self, feature_extractor: tf.keras.Model, vocab: str, d_model: int = 512, dff: int = 2048, num_heads: int = 8, # number of heads in the transformer decoder num_layers: int = 3, max_length: int = 50, dropout: float = 0.2, input_shape: Tuple[int, int, int] = (32, 128, 3), # different from the paper exportable: bool = False, cfg: Optional[Dict[str, Any]] = None, ) -> None: super().__init__() self.exportable = exportable self.max_length = max_length self.d_model = d_model self.vocab = vocab self.cfg = cfg self.vocab_size = len(vocab) self.feat_extractor = feature_extractor self.positional_encoding = PositionalEncoding(self.d_model, dropout, max_len=input_shape[0] * input_shape[1]) self.decoder = Decoder( num_layers=num_layers, d_model=self.d_model, num_heads=num_heads, vocab_size=self.vocab_size + 3, # EOS, SOS, PAD dff=dff, dropout=dropout, maximum_position_encoding=self.max_length, ) self.linear = layers.Dense(self.vocab_size + 3, kernel_initializer=tf.initializers.he_uniform()) self.postprocessor = MASTERPostProcessor(vocab=self.vocab) @tf.function def make_source_and_target_mask(self, source: tf.Tensor, target: tf.Tensor) -> Tuple[tf.Tensor, tf.Tensor]: # [1, 1, 1, ..., 0, 0, 0] -> 0 is masked # (N, 1, 1, max_length) target_pad_mask = tf.cast(tf.math.not_equal(target, self.vocab_size + 2), dtype=tf.uint8) target_pad_mask = target_pad_mask[:, tf.newaxis, tf.newaxis, :] target_length = target.shape[1] # sub mask filled diagonal with 1 = see 0 = masked (max_length, max_length) target_sub_mask = tf.linalg.band_part(tf.ones((target_length, target_length)), -1, 0) # source mask filled with ones (max_length, positional_encoded_seq_len) source_mask = tf.ones((target_length, source.shape[1])) # combine the two masks into one boolean mask where False is masked (N, 1, max_length, max_length) target_mask = tf.math.logical_and( tf.cast(target_sub_mask, dtype=tf.bool), tf.cast(target_pad_mask, dtype=tf.bool) ) return source_mask, target_mask @staticmethod def compute_loss( model_output: tf.Tensor, gt: tf.Tensor, seq_len: List[int], ) -> tf.Tensor: """Compute categorical cross-entropy loss for the model. Sequences are masked after the EOS character. Args: ---- gt: the encoded tensor with gt labels model_output: predicted logits of the model seq_len: lengths of each gt word inside the batch Returns: ------- The loss of the model on the batch """ # Input length : number of timesteps input_len = tf.shape(model_output)[1] # Add one for additional token (sos disappear in shift!) seq_len = tf.cast(seq_len, tf.int32) + 1 # One-hot gt labels oh_gt = tf.one_hot(gt, depth=model_output.shape[2]) # Compute loss: don't forget to shift gt! Otherwise the model learns to output the gt[t-1]! # The "masked" first gt char is . Delete last logit of the model output. cce = tf.nn.softmax_cross_entropy_with_logits(oh_gt[:, 1:, :], model_output[:, :-1, :]) # Compute mask mask_values = tf.zeros_like(cce) mask_2d = tf.sequence_mask(seq_len, input_len - 1) # delete the last mask timestep as well masked_loss = tf.where(mask_2d, cce, mask_values) ce_loss = tf.math.divide(tf.reduce_sum(masked_loss, axis=1), tf.cast(seq_len, model_output.dtype)) return tf.expand_dims(ce_loss, axis=1) def call( self, x: tf.Tensor, target: Optional[List[str]] = None, return_model_output: bool = False, return_preds: bool = False, **kwargs: Any, ) -> Dict[str, Any]: """Call function for training Args: ---- x: images target: list of str labels return_model_output: if True, return logits return_preds: if True, decode logits **kwargs: keyword arguments passed to the decoder Returns: ------- A dictionnary containing eventually loss, logits and predictions. """ # Encode feature = self.feat_extractor(x, **kwargs) b, h, w, c = feature.get_shape() # (N, H, W, C) --> (N, H * W, C) feature = tf.reshape(feature, shape=(b, h * w, c)) # add positional encoding to features encoded = self.positional_encoding(feature, **kwargs) out: Dict[str, tf.Tensor] = {} if kwargs.get("training", False) and target is None: raise ValueError("Need to provide labels during training") if target is not None: # Compute target: tensor of gts and sequence lengths gt, seq_len = self.build_target(target) # Compute decoder masks source_mask, target_mask = self.make_source_and_target_mask(encoded, gt) # Compute logits output = self.decoder(gt, encoded, source_mask, target_mask, **kwargs) logits = self.linear(output, **kwargs) else: logits = self.decode(encoded, **kwargs) logits = _bf16_to_float32(logits) if self.exportable: out["logits"] = logits return out if target is not None: out["loss"] = self.compute_loss(logits, gt, seq_len) if return_model_output: out["out_map"] = logits if return_preds: out["preds"] = self.postprocessor(logits) return out @tf.function def decode(self, encoded: tf.Tensor, **kwargs: Any) -> tf.Tensor: """Decode function for prediction Args: ---- encoded: encoded features **kwargs: keyword arguments passed to the decoder Returns: ------- A Tuple of tf.Tensor: predictions, logits """ b = encoded.shape[0] start_symbol = tf.constant(self.vocab_size + 1, dtype=tf.int32) # SOS padding_symbol = tf.constant(self.vocab_size + 2, dtype=tf.int32) # PAD ys = tf.fill(dims=(b, self.max_length - 1), value=padding_symbol) start_vector = tf.fill(dims=(b, 1), value=start_symbol) ys = tf.concat([start_vector, ys], axis=-1) # Final dimension include EOS/SOS/PAD for i in range(self.max_length - 1): source_mask, target_mask = self.make_source_and_target_mask(encoded, ys) output = self.decoder(ys, encoded, source_mask, target_mask, **kwargs) logits = self.linear(output, **kwargs) prob = tf.nn.softmax(logits, axis=-1) next_token = tf.argmax(prob, axis=-1, output_type=ys.dtype) # update ys with the next token and ignore the first token (SOS) i_mesh, j_mesh = tf.meshgrid(tf.range(b), tf.range(self.max_length), indexing="ij") indices = tf.stack([i_mesh[:, i + 1], j_mesh[:, i + 1]], axis=1) ys = tf.tensor_scatter_nd_update(ys, indices, next_token[:, i]) # Shape (N, max_length, vocab_size + 1) return logits class MASTERPostProcessor(_MASTERPostProcessor): """Post processor for MASTER architectures Args: ---- vocab: string containing the ordered sequence of supported characters """ def __call__( self, logits: tf.Tensor, ) -> List[Tuple[str, float]]: # compute pred with argmax for attention models out_idxs = tf.math.argmax(logits, axis=2) # N x L probs = tf.gather(tf.nn.softmax(logits, axis=-1), out_idxs, axis=-1, batch_dims=2) # Take the minimum confidence of the sequence probs = tf.math.reduce_min(probs, axis=1) # decode raw output of the model with tf_label_to_idx out_idxs = tf.cast(out_idxs, dtype="int32") embedding = tf.constant(self._embedding, dtype=tf.string) decoded_strings_pred = tf.strings.reduce_join(inputs=tf.nn.embedding_lookup(embedding, out_idxs), axis=-1) decoded_strings_pred = tf.strings.split(decoded_strings_pred, "") decoded_strings_pred = tf.sparse.to_dense(decoded_strings_pred.to_sparse(), default_value="not valid")[:, 0] word_values = [word.decode() for word in decoded_strings_pred.numpy().tolist()] return list(zip(word_values, probs.numpy().clip(0, 1).tolist())) def _master(arch: str, pretrained: bool, backbone_fn, pretrained_backbone: bool = True, **kwargs: Any) -> MASTER: pretrained_backbone = pretrained_backbone and not pretrained # Patch the config _cfg = deepcopy(default_cfgs[arch]) _cfg["input_shape"] = kwargs.get("input_shape", _cfg["input_shape"]) _cfg["vocab"] = kwargs.get("vocab", _cfg["vocab"]) kwargs["vocab"] = _cfg["vocab"] kwargs["input_shape"] = _cfg["input_shape"] # Build the model model = MASTER( backbone_fn(pretrained=pretrained_backbone, input_shape=_cfg["input_shape"], include_top=False), cfg=_cfg, **kwargs, ) # Load pretrained parameters if pretrained: load_pretrained_params(model, default_cfgs[arch]["url"]) return model def master(pretrained: bool = False, **kwargs: Any) -> MASTER: """MASTER as described in paper: `_. >>> import tensorflow as tf >>> from doctr.models import master >>> model = master(pretrained=False) >>> input_tensor = tf.random.uniform(shape=[1, 32, 128, 3], maxval=1, dtype=tf.float32) >>> out = model(input_tensor) Args: ---- pretrained (bool): If True, returns a model pre-trained on our text recognition dataset **kwargs: keywoard arguments passed to the MASTER architecture Returns: ------- text recognition architecture """ return _master("master", pretrained, magc_resnet31, **kwargs)