Mayo commited on
Commit
15ade02
·
unverified ·
1 Parent(s): 46ba5e4

Add complete model card metadata

Browse files
Files changed (2) hide show
  1. README.md +42 -3
  2. config.json +19 -0
README.md CHANGED
@@ -1,3 +1,42 @@
1
- ---
2
- license: gpl-3.0
3
- ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: gpl-3.0
3
+ pipeline_tag: object-detection
4
+ tags:
5
+ - comic
6
+ - manga
7
+ - text-detection
8
+ - text-segmentation
9
+ - object-detection
10
+ - yolo
11
+ - dbnet
12
+ - safetensors
13
+ ---
14
+
15
+ # Comic Text Detector
16
+
17
+ A split SafeTensors conversion of the comic text detector from [`dmMaze/BallonsTranslator`](https://github.com/dmMaze/BallonsTranslator). It combines a YOLOv5 feature extractor, a UNet text mask head, and a DBNet-style line detection head.
18
+
19
+ ## Model details
20
+
21
+ - Task: comic text-region and text-line detection
22
+ - YOLO component: FP16 weights
23
+ - UNet and DBNet components: FP32 weights
24
+ - Outputs: text mask, line probability maps, and structured text blocks after postprocessing
25
+ - Format: SafeTensors state dictionaries with source tensor names preserved
26
+
27
+ ## Files
28
+
29
+ - `yolo-v5.safetensors`: YOLOv5 backbone, neck, and detection-head state
30
+ - `unet.safetensors`: text segmentation head
31
+ - `dbnet.safetensors`: differentiable-binarization line head
32
+ - `config.json`: component and runtime metadata
33
+
34
+ The three files exactly preserve the key sets, shapes, dtypes, and tensor values from their corresponding sections in the original `comictextdetector.pt` checkpoint.
35
+
36
+ ## Intended use and limitations
37
+
38
+ Use this model to identify text regions and individual lines before comic OCR. It is specialized for illustrated pages and may perform poorly on photographs, handwriting, unusual layouts, or very small text. Detection thresholds and postprocessing are part of the complete BallonsTranslator/Koharu pipeline; loading only the tensor files is not sufficient to reproduce final text blocks. No evaluation metrics or training dataset artifacts are included here.
39
+
40
+ ## License
41
+
42
+ GPL-3.0, following the BallonsTranslator implementation.
config.json ADDED
@@ -0,0 +1,19 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model_type": "comic-text-detector",
3
+ "task": "object-detection",
4
+ "components": {
5
+ "yolo_v5": {
6
+ "file": "yolo-v5.safetensors",
7
+ "dtype": "float16"
8
+ },
9
+ "unet": {
10
+ "file": "unet.safetensors",
11
+ "dtype": "float32"
12
+ },
13
+ "dbnet": {
14
+ "file": "dbnet.safetensors",
15
+ "dtype": "float32"
16
+ }
17
+ },
18
+ "source_repository": "https://github.com/dmMaze/BallonsTranslator"
19
+ }