Description
General model for table detection inspired by https://arxiv.org/abs/2004.12629
Predicted Entities
How to use
This modes is used by ImageTableDetector
binary_to_image = BinaryToImage()
binary_to_image.setImageType(ImageType.TYPE_3BYTE_BGR)
table_detector = ImageTableDetector
.pretrained("general_model_table_detection_v2", "en", "clinical/ocr")
.setInputCol("image")
.setOutputCol("table_regions")
pipeline = PipelineModel(stages=[
binary_to_image,
table_detector
])
var imgDf = spark.read.format("binaryFile").load(imagePath)
var bin2imTransformer = new BinaryToImage()
bin2imTransformer.setImageType(ImageType.TYPE_3BYTE_BGR)
val dataFrame = bin2imTransformer.transform(imgDf)
val tableDetector = ImageTableDetector
.pretrained("general_model_table_detection_v2", "en", "clinical/ocr")
.setInputCol("image")
.setOutputCol("table regions")
Model Information
| Model Name: | ocr_table_detection_general_model |
| Type: | ocr |
| Compatibility: | Visual NLP 3.0.0+ |
| License: | Licensed |
| Edition: | Official |
| Output Labels: | [table regions] |
| Language: | en |