From 435fbdd5b6ff0214549609574e899ad428983a7b Mon Sep 17 00:00:00 2001
From: Manon blanco <blanco@teklia.com>
Date: Tue, 13 Jun 2023 10:01:22 +0000
Subject: [PATCH] Remove width_divisor and height_divisor parameters from
training configuration
---
dan/manager/ocr.py | 5 ++---
dan/ocr/document/train.py | 2 --
docs/usage/train/parameters.md | 2 --
tests/conftest.py | 2 --
4 files changed, 2 insertions(+), 9 deletions(-)
diff --git a/dan/manager/ocr.py b/dan/manager/ocr.py
index 533d074c..fdcd54a0 100644
--- a/dan/manager/ocr.py
+++ b/dan/manager/ocr.py
@@ -77,9 +77,8 @@ class OCRDataset(GenericDataset):
super(OCRDataset, self).__init__(params, set_name, custom_name, paths_and_sets)
self.charset = None
self.tokens = None
- self.reduce_dims_factor = np.array(
- [params["config"]["height_divisor"], params["config"]["width_divisor"], 1]
- )
+ # Factor to reduce the height and width of the feature vector before feeding the decoder.
+ self.reduce_dims_factor = np.array([32, 8, 1])
self.collate_function = OCRCollateFunction
def __getitem__(self, idx):
diff --git a/dan/ocr/document/train.py b/dan/ocr/document/train.py
index d0a5fc47..b1ae6bb5 100644
--- a/dan/ocr/document/train.py
+++ b/dan/ocr/document/train.py
@@ -105,8 +105,6 @@ def get_config():
"config": {
"load_in_memory": True, # Load all images in CPU memory
"worker_per_gpu": 4, # Num of parallel processes per gpu for data loading
- "width_divisor": 8, # Image width will be divided by 8
- "height_divisor": 32, # Image height will be divided by 32
"padding_value": 0, # Image padding value
"padding_token": None, # Label padding value
"preprocessings": [
diff --git a/docs/usage/train/parameters.md b/docs/usage/train/parameters.md
index 8d97ae63..03d334f4 100644
--- a/docs/usage/train/parameters.md
+++ b/docs/usage/train/parameters.md
@@ -14,8 +14,6 @@ All hyperparameters are specified and editable in the training scripts (meaning
| `dataset_params.config.datasets` | Dataset dictionary with the dataset name as key and dataset path as value. | `dict` | |
| `dataset_params.config.load_in_memory` | Load all images in CPU memory. | `str` | `True` |
| `dataset_params.config.worker_per_gpu` | Number of parallel processes per gpu for data loading. | `int` | `4` |
-| `dataset_params.config.height_divisor` | Factor to reduce the width of the feature vector before feeding the decoder. | `int` | `8` |
-| `dataset_params.config.width_divisor` | Factor to reduce the height of the feature vector before feeding the decoder. | `int` | `32` |
| `dataset_params.config.padding_value` | Image padding value. | `int` | `0` |
| `dataset_params.config.padding_token` | Transcription padding value. | `int` | `None` |
| `dataset_params.config.preprocessings` | List of pre-processing functions to apply to input images. | `list` | (see [dedicated section](#data-preprocessing)) |
diff --git a/tests/conftest.py b/tests/conftest.py
index e660869c..4d345847 100644
--- a/tests/conftest.py
+++ b/tests/conftest.py
@@ -68,8 +68,6 @@ def training_config():
},
"config": {
"load_in_memory": True, # Load all images in CPU memory
- "width_divisor": 8, # Image width will be divided by 8
- "height_divisor": 32, # Image height will be divided by 32
"padding_value": 0, # Image padding value
"padding_token": None, # Label padding value
"preprocessings": [
--
GitLab