python-doctr 0.11.0__tar.gz → 0.12.0__tar.gz

{python_doctr-0.11.0 → python_doctr-0.12.0}/PKG-INFO

@@ -1,6 +1,6 @@
-Metadata-Version: 2.2
+Metadata-Version: 2.4
 Name: python-doctr
-Version: 0.11.0
+Version: 0.12.0
 Summary: Document Text Recognition (docTR): deep Learning for high-performance OCR on documents.
 Author-email: Mindee <contact@mindee.com>
 Maintainer: François-Guillaume Fernandez, Charles Gaillard, Olivier Dulcy, Felix Dittrich
@@ -239,6 +239,7 @@ Requires-Dist: huggingface-hub<1.0.0,>=0.20.0
 Requires-Dist: Pillow>=9.2.0
 Requires-Dist: defusedxml>=0.7.0
 Requires-Dist: anyascii>=0.3.2
+Requires-Dist: validators>=0.18.0
 Requires-Dist: tqdm>=4.30.0
 Provides-Extra: tf
 Requires-Dist: tensorflow[and-cuda]<3.0.0,>=2.15.0; sys_platform == "linux" and extra == "tf"
@@ -302,12 +303,13 @@ Requires-Dist: recommonmark>=0.7.1; extra == "dev"
 Requires-Dist: sphinx-markdown-tables>=0.0.15; extra == "dev"
 Requires-Dist: sphinx-tabs>=3.3.0; extra == "dev"
 Requires-Dist: furo>=2022.3.4; extra == "dev"
+Dynamic: license-file
 
 <p align="center">
   <img src="https://github.com/mindee/doctr/raw/main/docs/images/Logo_doctr.gif" width="40%">
 </p>
 
-[![Slack Icon](https://img.shields.io/badge/Slack-Community-4A154B?style=flat-square&logo=slack&logoColor=white)](https://slack.mindee.com) [![License](https://img.shields.io/badge/License-Apache%202.0-blue.svg)](LICENSE) ![Build Status](https://github.com/mindee/doctr/workflows/builds/badge.svg) [![Docker Images](https://img.shields.io/badge/Docker-4287f5?style=flat&logo=docker&logoColor=white)](https://github.com/mindee/doctr/pkgs/container/doctr) [![codecov](https://codecov.io/gh/mindee/doctr/branch/main/graph/badge.svg?token=577MO567NM)](https://codecov.io/gh/mindee/doctr) [![CodeFactor](https://www.codefactor.io/repository/github/mindee/doctr/badge?s=bae07db86bb079ce9d6542315b8c6e70fa708a7e)](https://www.codefactor.io/repository/github/mindee/doctr) [![Codacy Badge](https://api.codacy.com/project/badge/Grade/340a76749b634586a498e1c0ab998f08)](https://app.codacy.com/gh/mindee/doctr?utm_source=github.com&utm_medium=referral&utm_content=mindee/doctr&utm_campaign=Badge_Grade) [![Doc Status](https://github.com/mindee/doctr/workflows/doc-status/badge.svg)](https://mindee.github.io/doctr) [![Pypi](https://img.shields.io/badge/pypi-v0.11.0-blue.svg)](https://pypi.org/project/python-doctr/) [![Hugging Face Spaces](https://img.shields.io/badge/%F0%9F%A4%97%20Hugging%20Face-Spaces-blue)](https://huggingface.co/spaces/mindee/doctr) [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/mindee/notebooks/blob/main/doctr/quicktour.ipynb) [![Gurubase](https://img.shields.io/badge/Gurubase-Ask%20docTR%20Guru-006BFF)](https://gurubase.io/g/doctr)
+[![Slack Icon](https://img.shields.io/badge/Slack-Community-4A154B?style=flat-square&logo=slack&logoColor=white)](https://slack.mindee.com) [![License](https://img.shields.io/badge/License-Apache%202.0-blue.svg)](LICENSE) ![Build Status](https://github.com/mindee/doctr/workflows/builds/badge.svg) [![Docker Images](https://img.shields.io/badge/Docker-4287f5?style=flat&logo=docker&logoColor=white)](https://github.com/mindee/doctr/pkgs/container/doctr) [![codecov](https://codecov.io/gh/mindee/doctr/branch/main/graph/badge.svg?token=577MO567NM)](https://codecov.io/gh/mindee/doctr) [![CodeFactor](https://www.codefactor.io/repository/github/mindee/doctr/badge?s=bae07db86bb079ce9d6542315b8c6e70fa708a7e)](https://www.codefactor.io/repository/github/mindee/doctr) [![Codacy Badge](https://api.codacy.com/project/badge/Grade/340a76749b634586a498e1c0ab998f08)](https://app.codacy.com/gh/mindee/doctr?utm_source=github.com&utm_medium=referral&utm_content=mindee/doctr&utm_campaign=Badge_Grade) [![Doc Status](https://github.com/mindee/doctr/workflows/doc-status/badge.svg)](https://mindee.github.io/doctr) [![Pypi](https://img.shields.io/badge/pypi-v0.12.0-blue.svg)](https://pypi.org/project/python-doctr/) [![Hugging Face Spaces](https://img.shields.io/badge/%F0%9F%A4%97%20Hugging%20Face-Spaces-blue)](https://huggingface.co/spaces/mindee/doctr) [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/mindee/notebooks/blob/main/doctr/quicktour.ipynb) [![Gurubase](https://img.shields.io/badge/Gurubase-Ask%20docTR%20Guru-006BFF)](https://gurubase.io/g/doctr)
 
 
 **Optical Character Recognition made seamless & accessible to anyone, powered by TensorFlow 2 & PyTorch**
@@ -438,6 +440,19 @@ The KIE predictor results per page are in a dictionary format with each key repr
 
 ## Installation
 
+> [!WARNING]
+> **TensorFlow Backend Deprecation Notice**
+>
+> Using docTR with TensorFlow as a backend is deprecated and will be removed in the next major release (v1.0.0).
+> We **recommend switching to the PyTorch backend**, which is more actively maintained and supports the latest features and models.
+> Alternatively, you can use [OnnxTR](https://github.com/felixdittrich92/OnnxTR), which does **not** require TensorFlow or PyTorch.
+>
+> This decision was made based on several considerations:
+>
+> - Allows better focus on improving the core library
+> - Frees up resources to develop new features faster
+> - Enables more targeted optimizations with PyTorch
+
 ### Prerequisites
 
 Python 3.10 (or higher) and [pip](https://pip.pypa.io/en/stable/) are required to install docTR.
@@ -504,6 +519,7 @@ Credits where it's due: this repository is implementing, among others, architect
 - MASTER: [MASTER: Multi-Aspect Non-local Network for Scene Text Recognition](https://arxiv.org/pdf/1910.02562.pdf).
 - ViTSTR: [Vision Transformer for Fast and Efficient Scene Text Recognition](https://arxiv.org/pdf/2105.08582.pdf).
 - PARSeq: [Scene Text Recognition with Permuted Autoregressive Sequence Models](https://arxiv.org/pdf/2207.06966).
+- VIPTR: [A Vision Permutable Extractor for Fast and Efficient Scene Text Recognition](https://arxiv.org/abs/2401.10110).
 
 ## More goodies
 

{python_doctr-0.11.0 → python_doctr-0.12.0}/README.md

@@ -2,7 +2,7 @@
   <img src="https://github.com/mindee/doctr/raw/main/docs/images/Logo_doctr.gif" width="40%">
 </p>
 
-[![Slack Icon](https://img.shields.io/badge/Slack-Community-4A154B?style=flat-square&logo=slack&logoColor=white)](https://slack.mindee.com) [![License](https://img.shields.io/badge/License-Apache%202.0-blue.svg)](LICENSE) ![Build Status](https://github.com/mindee/doctr/workflows/builds/badge.svg) [![Docker Images](https://img.shields.io/badge/Docker-4287f5?style=flat&logo=docker&logoColor=white)](https://github.com/mindee/doctr/pkgs/container/doctr) [![codecov](https://codecov.io/gh/mindee/doctr/branch/main/graph/badge.svg?token=577MO567NM)](https://codecov.io/gh/mindee/doctr) [![CodeFactor](https://www.codefactor.io/repository/github/mindee/doctr/badge?s=bae07db86bb079ce9d6542315b8c6e70fa708a7e)](https://www.codefactor.io/repository/github/mindee/doctr) [![Codacy Badge](https://api.codacy.com/project/badge/Grade/340a76749b634586a498e1c0ab998f08)](https://app.codacy.com/gh/mindee/doctr?utm_source=github.com&utm_medium=referral&utm_content=mindee/doctr&utm_campaign=Badge_Grade) [![Doc Status](https://github.com/mindee/doctr/workflows/doc-status/badge.svg)](https://mindee.github.io/doctr) [![Pypi](https://img.shields.io/badge/pypi-v0.11.0-blue.svg)](https://pypi.org/project/python-doctr/) [![Hugging Face Spaces](https://img.shields.io/badge/%F0%9F%A4%97%20Hugging%20Face-Spaces-blue)](https://huggingface.co/spaces/mindee/doctr) [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/mindee/notebooks/blob/main/doctr/quicktour.ipynb) [![Gurubase](https://img.shields.io/badge/Gurubase-Ask%20docTR%20Guru-006BFF)](https://gurubase.io/g/doctr)
+[![Slack Icon](https://img.shields.io/badge/Slack-Community-4A154B?style=flat-square&logo=slack&logoColor=white)](https://slack.mindee.com) [![License](https://img.shields.io/badge/License-Apache%202.0-blue.svg)](LICENSE) ![Build Status](https://github.com/mindee/doctr/workflows/builds/badge.svg) [![Docker Images](https://img.shields.io/badge/Docker-4287f5?style=flat&logo=docker&logoColor=white)](https://github.com/mindee/doctr/pkgs/container/doctr) [![codecov](https://codecov.io/gh/mindee/doctr/branch/main/graph/badge.svg?token=577MO567NM)](https://codecov.io/gh/mindee/doctr) [![CodeFactor](https://www.codefactor.io/repository/github/mindee/doctr/badge?s=bae07db86bb079ce9d6542315b8c6e70fa708a7e)](https://www.codefactor.io/repository/github/mindee/doctr) [![Codacy Badge](https://api.codacy.com/project/badge/Grade/340a76749b634586a498e1c0ab998f08)](https://app.codacy.com/gh/mindee/doctr?utm_source=github.com&utm_medium=referral&utm_content=mindee/doctr&utm_campaign=Badge_Grade) [![Doc Status](https://github.com/mindee/doctr/workflows/doc-status/badge.svg)](https://mindee.github.io/doctr) [![Pypi](https://img.shields.io/badge/pypi-v0.12.0-blue.svg)](https://pypi.org/project/python-doctr/) [![Hugging Face Spaces](https://img.shields.io/badge/%F0%9F%A4%97%20Hugging%20Face-Spaces-blue)](https://huggingface.co/spaces/mindee/doctr) [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/mindee/notebooks/blob/main/doctr/quicktour.ipynb) [![Gurubase](https://img.shields.io/badge/Gurubase-Ask%20docTR%20Guru-006BFF)](https://gurubase.io/g/doctr)
 
 
 **Optical Character Recognition made seamless & accessible to anyone, powered by TensorFlow 2 & PyTorch**
@@ -133,6 +133,19 @@ The KIE predictor results per page are in a dictionary format with each key repr
 
 ## Installation
 
+> [!WARNING]
+> **TensorFlow Backend Deprecation Notice**
+>
+> Using docTR with TensorFlow as a backend is deprecated and will be removed in the next major release (v1.0.0).
+> We **recommend switching to the PyTorch backend**, which is more actively maintained and supports the latest features and models.
+> Alternatively, you can use [OnnxTR](https://github.com/felixdittrich92/OnnxTR), which does **not** require TensorFlow or PyTorch.
+>
+> This decision was made based on several considerations:
+>
+> - Allows better focus on improving the core library
+> - Frees up resources to develop new features faster
+> - Enables more targeted optimizations with PyTorch
+
 ### Prerequisites
 
 Python 3.10 (or higher) and [pip](https://pip.pypa.io/en/stable/) are required to install docTR.
@@ -199,6 +212,7 @@ Credits where it's due: this repository is implementing, among others, architect
 - MASTER: [MASTER: Multi-Aspect Non-local Network for Scene Text Recognition](https://arxiv.org/pdf/1910.02562.pdf).
 - ViTSTR: [Vision Transformer for Fast and Efficient Scene Text Recognition](https://arxiv.org/pdf/2105.08582.pdf).
 - PARSeq: [Scene Text Recognition with Permuted Autoregressive Sequence Models](https://arxiv.org/pdf/2207.06966).
+- VIPTR: [A Vision Permutable Extractor for Fast and Efficient Scene Text Recognition](https://arxiv.org/abs/2401.10110).
 
 ## More goodies
 

{python_doctr-0.11.0 → python_doctr-0.12.0}/doctr/datasets/__init__.py

@@ -1,6 +1,7 @@
 from doctr.file_utils import is_tf_available
 
 from .generator import *
+from .coco_text import *
 from .cord import *
 from .detection import *
 from .doc_artefacts import *

python_doctr-0.12.0/doctr/datasets/coco_text.py

@@ -0,0 +1,139 @@
+# Copyright (C) 2021-2025, Mindee.
+
+# This program is licensed under the Apache License 2.0.
+# See LICENSE or go to <https://opensource.org/licenses/Apache-2.0> for full license details.
+
+import json
+import os
+from pathlib import Path
+from typing import Any
+
+import numpy as np
+from tqdm import tqdm
+
+from .datasets import AbstractDataset
+from .utils import convert_target_to_relative, crop_bboxes_from_image
+
+__all__ = ["COCOTEXT"]
+
+
+class COCOTEXT(AbstractDataset):
+    """
+    COCO-Text dataset from `"COCO-Text: Dataset and Benchmark for Text Detection and Recognition in Natural Images"
+    <https://arxiv.org/pdf/1601.07140v2>`_ |
+    `"homepage" <https://bgshih.github.io/cocotext/>`_.
+
+    >>> # NOTE: You need to download the dataset first.
+    >>> from doctr.datasets import COCOTEXT
+    >>> train_set = COCOTEXT(train=True, img_folder="/path/to/coco_text/train2014/",
+    >>>                     label_path="/path/to/coco_text/cocotext.v2.json")
+    >>> img, target = train_set[0]
+    >>> test_set = COCOTEXT(train=False, img_folder="/path/to/coco_text/train2014/",
+    >>> label_path = "/path/to/coco_text/cocotext.v2.json")
+    >>> img, target = test_set[0]
+
+    Args:
+        img_folder: folder with all the images of the dataset
+        label_path: path to the annotations file of the dataset
+        train: whether the subset should be the training one
+        use_polygons: whether polygons should be considered as rotated bounding box (instead of straight ones)
+        recognition_task: whether the dataset should be used for recognition task
+        detection_task: whether the dataset should be used for detection task
+        **kwargs: keyword arguments from `AbstractDataset`.
+    """
+
+    def __init__(
+        self,
+        img_folder: str,
+        label_path: str,
+        train: bool = True,
+        use_polygons: bool = False,
+        recognition_task: bool = False,
+        detection_task: bool = False,
+        **kwargs: Any,
+    ) -> None:
+        super().__init__(
+            img_folder, pre_transforms=convert_target_to_relative if not recognition_task else None, **kwargs
+        )
+        # Task check
+        if recognition_task and detection_task:
+            raise ValueError(
+                " 'recognition' and 'detection task' cannot be set to True simultaneously. "
+                + " To get the whole dataset with boxes and labels leave both parameters to False "
+            )
+
+        # File existence check
+        if not os.path.exists(label_path) or not os.path.exists(img_folder):
+            raise FileNotFoundError(f"unable to find {label_path if not os.path.exists(label_path) else img_folder}")
+
+        tmp_root = img_folder
+        self.train = train
+        np_dtype = np.float32
+        self.data: list[tuple[str | Path | np.ndarray, str | dict[str, Any] | np.ndarray]] = []
+
+        with open(label_path, "r") as file:
+            data = json.load(file)
+
+        # Filter images based on the set
+        img_items = [img for img in data["imgs"].items() if (img[1]["set"] == "train") == train]
+        box: list[float] | np.ndarray
+
+        for img_id, img_info in tqdm(img_items, desc="Preparing and Loading COCOTEXT", total=len(img_items)):
+            img_path = os.path.join(img_folder, img_info["file_name"])
+
+            # File existence check
+            if not os.path.exists(img_path):  # pragma: no cover
+                raise FileNotFoundError(f"Unable to locate {img_path}")
+
+            # Get annotations for the current image (only legible text)
+            annotations = [
+                ann
+                for ann in data["anns"].values()
+                if ann["image_id"] == int(img_id) and ann["legibility"] == "legible"
+            ]
+
+            # Some images have no annotations with readable text
+            if not annotations:  # pragma: no cover
+                continue
+
+            _targets = []
+
+            for annotation in annotations:
+                x, y, w, h = annotation["bbox"]
+                if use_polygons:
+                    # (x, y) coordinates of top left, top right, bottom right, bottom left corners
+                    box = np.array(
+                        [
+                            [x, y],
+                            [x + w, y],
+                            [x + w, y + h],
+                            [x, y + h],
+                        ],
+                        dtype=np_dtype,
+                    )
+                else:
+                    # (xmin, ymin, xmax, ymax) coordinates
+                    box = [x, y, x + w, y + h]
+                _targets.append((annotation["utf8_string"], box))
+            text_targets, box_targets = zip(*_targets)
+
+            if recognition_task:
+                crops = crop_bboxes_from_image(
+                    img_path=os.path.join(tmp_root, img_path), geoms=np.asarray(box_targets, dtype=int).clip(min=0)
+                )
+                for crop, label in zip(crops, list(text_targets)):
+                    if label and " " not in label:
+                        self.data.append((crop, label))
+
+            elif detection_task:
+                self.data.append((img_path, np.asarray(box_targets, dtype=int).clip(min=0)))
+            else:
+                self.data.append((
+                    img_path,
+                    dict(boxes=np.asarray(box_targets, dtype=int).clip(min=0), labels=list(text_targets)),
+                ))
+
+        self.root = tmp_root
+
+    def extra_repr(self) -> str:
+        return f"train={self.train}"

{python_doctr-0.11.0 → python_doctr-0.12.0}/doctr/datasets/cord.py

@@ -116,7 +116,8 @@ class CORD(VisionDataset):
                     img_path=os.path.join(tmp_root, img_path), geoms=np.asarray(box_targets, dtype=int).clip(min=0)
                 )
                 for crop, label in zip(crops, list(text_targets)):
-                    self.data.append((crop, label))
+                    if " " not in label:
+                        self.data.append((crop, label))
             elif detection_task:
                 self.data.append((img_path, np.asarray(box_targets, dtype=int).clip(min=0)))
             else:

{python_doctr-0.11.0 → python_doctr-0.12.0}/doctr/datasets/funsd.py

@@ -107,8 +107,8 @@ class FUNSD(VisionDataset):
                 )
                 for crop, label in zip(crops, list(text_targets)):
                     # filter labels with unknown characters
-                    if not any(char in label for char in ["☑", "☐", "\uf703", "\uf702"]):
-                        self.data.append((crop, label))
+                    if not any(char in label for char in ["☑", "☐", "\u03bf", "\uf703", "\uf702", " "]):
+                        self.data.append((crop, label.replace("–", "-")))
             elif detection_task:
                 self.data.append((img_path, np.asarray(box_targets, dtype=np_dtype)))
             else:

{python_doctr-0.11.0 → python_doctr-0.12.0}/doctr/datasets/ic03.py

@@ -122,7 +122,7 @@ class IC03(VisionDataset):
                 if recognition_task:
                     crops = crop_bboxes_from_image(img_path=os.path.join(tmp_root, name.text), geoms=boxes)
                     for crop, label in zip(crops, labels):
-                        if crop.shape[0] > 0 and crop.shape[1] > 0 and len(label) > 0:
+                        if crop.shape[0] > 0 and crop.shape[1] > 0 and len(label) > 0 and " " not in label:
                             self.data.append((crop, label))
                 elif detection_task:
                     self.data.append((name.text, boxes))

{python_doctr-0.11.0 → python_doctr-0.12.0}/doctr/datasets/ic13.py

@@ -100,7 +100,8 @@ class IC13(AbstractDataset):
             if recognition_task:
                 crops = crop_bboxes_from_image(img_path=img_path, geoms=box_targets)
                 for crop, label in zip(crops, labels):
-                    self.data.append((crop, label))
+                    if " " not in label:
+                        self.data.append((crop, label))
             elif detection_task:
                 self.data.append((img_path, box_targets))
             else:

{python_doctr-0.11.0 → python_doctr-0.12.0}/doctr/datasets/iiit5k.py

@@ -8,6 +8,7 @@ from typing import Any
 
 import numpy as np
 import scipy.io as sio
+from PIL import Image
 from tqdm import tqdm
 
 from .datasets import VisionDataset
@@ -98,7 +99,9 @@ class IIIT5K(VisionDataset):
                 box_targets = [[box[0], box[1], box[0] + box[2], box[1] + box[3]] for box in box_targets]
 
             if recognition_task:
-                self.data.append((_raw_path, _raw_label))
+                if " " not in _raw_label:
+                    with Image.open(os.path.join(tmp_root, _raw_path)) as pil_img:
+                        self.data.append((np.array(pil_img.convert("RGB")), _raw_label))
             elif detection_task:
                 self.data.append((_raw_path, np.asarray(box_targets, dtype=np_dtype)))
             else:

{python_doctr-0.11.0 → python_doctr-0.12.0}/doctr/datasets/imgur5k.py

@@ -133,7 +133,13 @@ class IMGUR5K(AbstractDataset):
                         img_path=os.path.join(self.root, img_name), geoms=np.asarray(box_targets, dtype=np_dtype)
                     )
                     for crop, label in zip(crops, labels):
-                        if crop.shape[0] > 0 and crop.shape[1] > 0 and len(label) > 0:
+                        if (
+                            crop.shape[0] > 0
+                            and crop.shape[1] > 0
+                            and len(label) > 0
+                            and len(label) < 30
+                            and " " not in label
+                        ):
                             # write data to disk
                             with open(os.path.join(reco_folder_path, f"{reco_images_counter}.txt"), "w") as f:
                                 f.write(label)
@@ -152,6 +158,7 @@ class IMGUR5K(AbstractDataset):
         return f"train={self.train}"
 
     def _read_from_folder(self, path: str) -> None:
-        for img_path in glob.glob(os.path.join(path, "*.png")):
+        img_paths = glob.glob(os.path.join(path, "*.png"))
+        for img_path in tqdm(iterable=img_paths, desc="Preparing and Loading IMGUR5K", total=len(img_paths)):
             with open(os.path.join(path, f"{os.path.basename(img_path)[:-4]}.txt"), "r") as f:
                 self.data.append((img_path, f.read()))

{python_doctr-0.11.0 → python_doctr-0.12.0}/doctr/datasets/loader.py

@@ -19,7 +19,7 @@ def default_collate(samples):
         samples: list of N tuples containing M elements
 
     Returns:
-        tuple of M sequences contianing N elements each
+        tuple of M sequences containing N elements each
     """
     batch_data = zip(*samples)
 

{python_doctr-0.11.0 → python_doctr-0.12.0}/doctr/datasets/ocr.py

@@ -40,7 +40,7 @@ class OCRDataset(AbstractDataset):
         super().__init__(img_folder, **kwargs)
 
         # List images
-        self.data: list[tuple[str, dict[str, Any]]] = []
+        self.data: list[tuple[Path, dict[str, Any]]] = []
         np_dtype = np.float32
         with open(label_file, "rb") as f:
             data = json.load(f)

{python_doctr-0.11.0 → python_doctr-0.12.0}/doctr/datasets/recognition.py

@@ -23,7 +23,7 @@ class RecognitionDataset(AbstractDataset):
 
     Args:
         img_folder: path to the images folder
-        labels_path: pathe to the json file containing all labels (character sequences)
+        labels_path: path to the json file containing all labels (character sequences)
         **kwargs: keyword arguments from `AbstractDataset`.
     """
 

{python_doctr-0.11.0 → python_doctr-0.12.0}/doctr/datasets/svhn.py

@@ -129,7 +129,7 @@ class SVHN(VisionDataset):
                 if recognition_task:
                     crops = crop_bboxes_from_image(img_path=os.path.join(tmp_root, img_name), geoms=box_targets)
                     for crop, label in zip(crops, label_targets):
-                        if crop.shape[0] > 0 and crop.shape[1] > 0 and len(label) > 0:
+                        if crop.shape[0] > 0 and crop.shape[1] > 0 and len(label) > 0 and " " not in label:
                             self.data.append((crop, label))
                 elif detection_task:
                     self.data.append((img_name, box_targets))

{python_doctr-0.11.0 → python_doctr-0.12.0}/doctr/datasets/svt.py

@@ -35,7 +35,7 @@ class SVT(VisionDataset):
         **kwargs: keyword arguments from `VisionDataset`.
     """
 
-    URL = "http://vision.ucsd.edu/~kai/svt/svt.zip"
+    URL = "http://www.iapr-tc11.org/dataset/SVT/svt.zip"
     SHA256 = "63b3d55e6b6d1e036e2a844a20c034fe3af3c32e4d914d6e0c4a3cd43df3bebf"
 
     def __init__(
@@ -113,7 +113,7 @@ class SVT(VisionDataset):
             if recognition_task:
                 crops = crop_bboxes_from_image(img_path=os.path.join(tmp_root, name.text), geoms=boxes)
                 for crop, label in zip(crops, labels):
-                    if crop.shape[0] > 0 and crop.shape[1] > 0 and len(label) > 0:
+                    if crop.shape[0] > 0 and crop.shape[1] > 0 and len(label) > 0 and " " not in label:
                         self.data.append((crop, label))
             elif detection_task:
                 self.data.append((name.text, boxes))

{python_doctr-0.11.0 → python_doctr-0.12.0}/doctr/datasets/synthtext.py

@@ -41,6 +41,12 @@ class SynthText(VisionDataset):
     URL = "https://thor.robots.ox.ac.uk/~vgg/data/scenetext/SynthText.zip"
     SHA256 = "28ab030485ec8df3ed612c568dd71fb2793b9afbfa3a9d9c6e792aef33265bf1"
 
+    # filter corrupted or missing images
+    BLACKLIST = (
+        "67/fruits_129_",
+        "194/window_19_",
+    )
+
     def __init__(
         self,
         train: bool = True,
@@ -111,7 +117,13 @@ class SynthText(VisionDataset):
             if recognition_task:
                 crops = crop_bboxes_from_image(img_path=os.path.join(tmp_root, img_path[0]), geoms=word_boxes)
                 for crop, label in zip(crops, labels):
-                    if crop.shape[0] > 0 and crop.shape[1] > 0 and len(label) > 0:
+                    if (
+                        crop.shape[0] > 0
+                        and crop.shape[1] > 0
+                        and len(label) > 0
+                        and len(label) < 30
+                        and " " not in label
+                    ):
                         # write data to disk
                         with open(os.path.join(reco_folder_path, f"{reco_images_counter}.txt"), "w") as f:
                             f.write(label)
@@ -132,6 +144,7 @@ class SynthText(VisionDataset):
         return f"train={self.train}"
 
     def _read_from_folder(self, path: str) -> None:
-        for img_path in glob.glob(os.path.join(path, "*.png")):
+        img_paths = glob.glob(os.path.join(path, "*.png"))
+        for img_path in tqdm(iterable=img_paths, desc="Preparing and Loading SynthText", total=len(img_paths)):
             with open(os.path.join(path, f"{os.path.basename(img_path)[:-4]}.txt"), "r") as f:
                 self.data.append((img_path, f.read()))

{python_doctr-0.11.0 → python_doctr-0.12.0}/doctr/datasets/utils.py

@@ -48,7 +48,7 @@ def translate(
         A string translated in a given vocab
     """
     if VOCABS.get(vocab_name) is None:
-        raise KeyError("output vocabulary must be in vocabs dictionnary")
+        raise KeyError("output vocabulary must be in vocabs dictionary")
 
     translated = ""
     for char in input_string:
@@ -81,11 +81,12 @@ def encode_string(
     """
     try:
         return list(map(vocab.index, input_string))
-    except ValueError:
+    except ValueError as e:
+        missing_chars = [char for char in input_string if char not in vocab]
         raise ValueError(
-            f"some characters cannot be found in 'vocab'. \
-                         Please check the input string {input_string} and the vocabulary {vocab}"
-        )
+            f"Some characters cannot be found in 'vocab': {set(missing_chars)}.\n"
+            f"Please check the input string `{input_string}` and the vocabulary `{vocab}`"
+        ) from e
 
 
 def decode_sequence(
@@ -199,7 +200,7 @@ def crop_bboxes_from_image(img_path: str | Path, geoms: np.ndarray) -> list[np.n
         a list of cropped images
     """
     with Image.open(img_path) as pil_img:
-        img: np.ndarray = np.array(pil_img.convert("RGB"))
+        img: np.ndarray = np.asarray(pil_img.convert("RGB"))
     # Polygon
     if geoms.ndim == 3 and geoms.shape[1:] == (4, 2):
         return extract_rcrops(img, geoms.astype(dtype=int))