from 20260830-030109_syn-real-v2

README.md

@@ -0,0 +1,61 @@
+---
+license: mit
+library_name: segmentation-models-pytorch
+pipeline_tag: image-segmentation
+tags:
+  - comics
+  - manga
+  - text-segmentation
+language:
+  - en
+  - ja
+  - ko
+  - zh
+---
+
+# comic-text-mask
+
+Binary **text-mask segmenter** for the
+[comic-localizer](https://github.com/TareHimself/comic-localizer) cleaning
+pipeline. It runs on a detector's text-region crop and returns a per-pixel
+"text vs not-text" mask that feeds LaMa inpainting. Locating and grouping text
+is the detector's job, not this model's.
+
+Training data (see the [training repo](https://github.com/TareHimself/comic-localizer-text-masking)): synthetic text rendered onto
+(a) procedurally generated flat surfaces with synthetic clutter and (b) real
+cleaned comic pages, framed as detector-style crops. Text is Latin, Japanese
+(kana + kanji, horizontal and vertical), Korean, and Chinese, with a large
+fraction of random-glyph runs so rare characters are covered. No source imagery
+is redistributed.
+
+Validation (held-out synthetic + real crops): IoU 0.918,
+precision 0.956, recall 0.959.
+
+## Use
+
+```python
+import numpy as np, torch
+from huggingface_hub import hf_hub_download
+
+meta = json.load(open(hf_hub_download("TareHimself/comic-text-mask", "tm_meta.json")))
+model = torch.jit.load(hf_hub_download("TareHimself/comic-text-mask", "model.pt")).eval()
+S = meta["imgsz"]
+
+# letterbox `rgb` (H,W,3 uint8) into an SxS square, pad 0, keep the paste box
+# ... then:
+x = torch.from_numpy(square).permute(2, 0, 1).unsqueeze(0)      # (1,3,S,S) uint8
+prob = model(x)[0, 0].numpy()                                   # (S,S) float
+mask = (prob > meta["threshold"]).astype("uint8") * 255
+# crop the paste box back out and resize to the original size
+```
+
+Or load the raw weights with `segmentation-models-pytorch`:
+
+```python
+import segmentation_models_pytorch as smp
+model = smp.from_pretrained("TareHimself/comic-text-mask")   # normalisation NOT baked in
+```
+
+`model.pt` has `/255`, ImageNet normalisation, and the final sigmoid baked into
+the graph; it expects letterboxed **uint8** RGB. `model.safetensors` is the
+pristine network.

config.json

@@ -0,0 +1,20 @@
+{
+  "_model_class": "Unet",
+  "activation": null,
+  "aux_params": null,
+  "classes": 1,
+  "decoder_attention_type": null,
+  "decoder_channels": [
+    256,
+    128,
+    64,
+    32,
+    16
+  ],
+  "decoder_interpolation": "nearest",
+  "decoder_use_norm": "batchnorm",
+  "encoder_depth": 5,
+  "encoder_name": "resnet18",
+  "encoder_weights": null,
+  "in_channels": 3
+}

model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:37fa988fcf2908c23d75af3988b00fd3210027627052b12a5abf606ac2d65da7
+size 57586952

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0d1299ce73b89cf7065418f00b3d4806b8f73910299fd30c41e83a3463c4e185
+size 57377484

tm_meta.json

@@ -0,0 +1,22 @@
+{
+  "imgsz": 384,
+  "input": "letterboxed uint8 RGB (1,3,imgsz,imgsz), pad 0",
+  "output": "float32 text probability (1,1,imgsz,imgsz)",
+  "normalization": "baked into model.pt",
+  "threshold": 0.5,
+  "arch": "unet",
+  "encoder": "resnet18",
+  "metrics": {
+    "iou": 0.91798,
+    "precision": 0.95562,
+    "recall": 0.95885
+  },
+  "languages": [
+    "en_GB",
+    "en_US",
+    "ja",
+    "ko",
+    "zh_CN",
+    "zh_TW"
+  ]
+}