feat: init

.gitattributes

@@ -1,35 +1,50 @@
-*.7z filter=lfs diff=lfs merge=lfs -text
-*.arrow filter=lfs diff=lfs merge=lfs -text
-*.bin filter=lfs diff=lfs merge=lfs -text
-*.bz2 filter=lfs diff=lfs merge=lfs -text
-*.ckpt filter=lfs diff=lfs merge=lfs -text
-*.ftz filter=lfs diff=lfs merge=lfs -text
-*.gz filter=lfs diff=lfs merge=lfs -text
-*.h5 filter=lfs diff=lfs merge=lfs -text
-*.joblib filter=lfs diff=lfs merge=lfs -text
-*.lfs.* filter=lfs diff=lfs merge=lfs -text
-*.mlmodel filter=lfs diff=lfs merge=lfs -text
-*.model filter=lfs diff=lfs merge=lfs -text
-*.msgpack filter=lfs diff=lfs merge=lfs -text
-*.npy filter=lfs diff=lfs merge=lfs -text
-*.npz filter=lfs diff=lfs merge=lfs -text
-*.onnx filter=lfs diff=lfs merge=lfs -text
-*.ot filter=lfs diff=lfs merge=lfs -text
-*.parquet filter=lfs diff=lfs merge=lfs -text
-*.pb filter=lfs diff=lfs merge=lfs -text
-*.pickle filter=lfs diff=lfs merge=lfs -text
-*.pkl filter=lfs diff=lfs merge=lfs -text
-*.pt filter=lfs diff=lfs merge=lfs -text
-*.pth filter=lfs diff=lfs merge=lfs -text
-*.rar filter=lfs diff=lfs merge=lfs -text
-*.safetensors filter=lfs diff=lfs merge=lfs -text
-saved_model/**/* filter=lfs diff=lfs merge=lfs -text
-*.tar.* filter=lfs diff=lfs merge=lfs -text
-*.tar filter=lfs diff=lfs merge=lfs -text
-*.tflite filter=lfs diff=lfs merge=lfs -text
-*.tgz filter=lfs diff=lfs merge=lfs -text
-*.wasm filter=lfs diff=lfs merge=lfs -text
-*.xz filter=lfs diff=lfs merge=lfs -text
-*.zip filter=lfs diff=lfs merge=lfs -text
-*.zst filter=lfs diff=lfs merge=lfs -text
-*tfevents* filter=lfs diff=lfs merge=lfs -text
+*.7z filter=lfs diff=lfs merge=lfs -text
+*.arrow filter=lfs diff=lfs merge=lfs -text
+*.bin filter=lfs diff=lfs merge=lfs -text
+*.bin.* filter=lfs diff=lfs merge=lfs -text
+*.bz2 filter=lfs diff=lfs merge=lfs -text
+*.ftz filter=lfs diff=lfs merge=lfs -text
+*.gz filter=lfs diff=lfs merge=lfs -text
+*.h5 filter=lfs diff=lfs merge=lfs -text
+*.joblib filter=lfs diff=lfs merge=lfs -text
+*.lfs.* filter=lfs diff=lfs merge=lfs -text
+*.model filter=lfs diff=lfs merge=lfs -text
+*.msgpack filter=lfs diff=lfs merge=lfs -text
+*.onnx filter=lfs diff=lfs merge=lfs -text
+*.ot filter=lfs diff=lfs merge=lfs -text
+*.parquet filter=lfs diff=lfs merge=lfs -text
+*.pb filter=lfs diff=lfs merge=lfs -text
+ 
+*.pth filter=lfs diff=lfs merge=lfs -text
+*.rar filter=lfs diff=lfs merge=lfs -text
+saved_model/**/* filter=lfs diff=lfs merge=lfs -text
+*.tar.* filter=lfs diff=lfs merge=lfs -text
+*.tflite filter=lfs diff=lfs merge=lfs -text
+*.tgz filter=lfs diff=lfs merge=lfs -text
+*.xz filter=lfs diff=lfs merge=lfs -text
+*.zip filter=lfs diff=lfs merge=lfs -text
+*.zstandard filter=lfs diff=lfs merge=lfs -text
+*.tfevents* filter=lfs diff=lfs merge=lfs -text
+*.db* filter=lfs diff=lfs merge=lfs -text
+*.ark* filter=lfs diff=lfs merge=lfs -text
+**/*ckpt*data* filter=lfs diff=lfs merge=lfs -text
+**/*ckpt*.meta filter=lfs diff=lfs merge=lfs -text
+**/*ckpt*.index filter=lfs diff=lfs merge=lfs -text
+ 
+ 
+*.gguf* filter=lfs diff=lfs merge=lfs -text
+*.ggml filter=lfs diff=lfs merge=lfs -text
+*.llamafile* filter=lfs diff=lfs merge=lfs -text
+*.pt2 filter=lfs diff=lfs merge=lfs -text
+*.mlmodel filter=lfs diff=lfs merge=lfs -text
+*.npy filter=lfs diff=lfs merge=lfs -text
+*.npz filter=lfs diff=lfs merge=lfs -text
+*.pickle filter=lfs diff=lfs merge=lfs -text
+*.pkl filter=lfs diff=lfs merge=lfs -text
+*.tar filter=lfs diff=lfs merge=lfs -text
+*.wasm filter=lfs diff=lfs merge=lfs -text
+*.zst filter=lfs diff=lfs merge=lfs -text
+*tfevents* filter=lfs diff=lfs merge=lfs -text
+
+model.safetensors filter=lfs diff=lfs merge=lfs -text
+best.ckpt filter=lfs diff=lfs merge=lfs -text

.gitignore

@@ -0,0 +1,36 @@
+# Python
+__pycache__/
+*.py[oc]
+*.pyc
+*.pyo
+*.pyd
+
+# IDE
+.vscode/
+.idea/
+*.swp
+*.swo
+
+# OS
+.DS_Store
+Thumbs.db
+desktop.ini
+
+# Training outputs
+checkpoints/
+output/
+logs/
+tb_logs/
+wandb/
+runs/
+
+# Temporary files
+*.tmp
+*.temp
+
+# Secrets
+.env
+*.env
+secrets.json
+credentials.json
+api_keys.json

README.md

@@ -1,3 +1,116 @@
----
-license: apache-2.0
----
+# ink-eraser-latest（手写墨迹擦除模型）
+
+本目录是一个用于“手写墨迹擦除 / 文档去涂写”的模型导出包（Hugging Face 兼容格式）。模型输入为带墨迹的 RGB 图像，输出为去除墨迹后的 RGB 图像。
+
+## 模型信息
+
+- 架构：U-Net++（`segmentation-models-pytorch`）+ ResNet50 编码器
+- 任务：图像到图像（去除手写笔迹/墨迹）
+- 输入：RGB，形状 `[B, 3, H, W]`，数值范围 `[0, 1]`
+- 输出：RGB，形状 `[B, 3, H, W]`，数值范围 `[0, 1]`（末端 `sigmoid`）
+
+## 文件说明
+
+- `config.json`：模型结构与训练超参数（导出时写入）
+- `model.safetensors`：推理用权重（推荐）
+- `best.ckpt`：原始 PyTorch Lightning checkpoint（用于继续训练/复现实验）
+- `configuration.json`：简要元数据（framework/task）
+
+## 快速推理（SafeTensors，推荐）
+
+依赖：`torch`、`torchvision`、`segmentation-models-pytorch`、`safetensors`，以及 `Pillow`（读写图片可选）。
+
+```bash
+pip install torch torchvision segmentation-models-pytorch safetensors pillow
+```
+
+```python
+import json
+from pathlib import Path
+
+import torch
+import segmentation_models_pytorch as smp
+from safetensors.torch import load_file
+from PIL import Image
+import torchvision.transforms.functional as TF
+
+device = "cuda" if torch.cuda.is_available() else "cpu"
+
+# 1) 读取配置
+cfg = json.loads(Path("config.json").read_text(encoding="utf-8"))
+
+# 2) 构建网络（与导出配置保持一致）
+model = smp.UnetPlusPlus(
+    encoder_name=cfg["encoder_name"],
+    encoder_weights=None,  # 权重来自 model.safetensors
+    in_channels=cfg["in_channels"],
+    classes=cfg["classes"],
+    decoder_attention_type=cfg.get("decoder_attention_type"),
+    activation=cfg.get("activation"),  # 通常为 "sigmoid"
+).to(device)
+
+# 3) 加载权重
+# 说明：导出时可能混入非网络权重（例如 `edge_loss.kx/ky`），推理只需要 Unet++ 本体参数，过滤掉即可。
+state_dict = load_file("model.safetensors")
+model_keys = set(model.state_dict().keys())
+state_dict = {k: v for k, v in state_dict.items() if k in model_keys}
+model.load_state_dict(state_dict, strict=True)
+model.eval()
+
+# 4) 准备输入（训练时仅做 0~1 归一化；如需更贴近训练分布可 resize 到 512x512）
+img = Image.open("input.png").convert("RGB")
+x = TF.to_tensor(img).unsqueeze(0).to(device)  # [1,3,H,W] in [0,1]
+
+with torch.no_grad():
+    y = model(x).clamp(0, 1)  # [1,3,H,W]
+
+out = TF.to_pil_image(y.squeeze(0).cpu())
+out.save("output.png")
+```
+
+提示：若输入尺寸不是 32 的倍数，部分编码器结构可能要求先 `pad/resize` 到合适尺寸（例如 `512x512`）。
+
+也可以直接使用本项目提供的高清切块推理脚本（自动对大图切块并融合回原图），从项目根目录运行：
+
+```bash
+python infer_hd.py --model-dir assets/InkErase --input input.png --output output.png
+```
+
+## 使用 `best.ckpt`（继续训练/复现实验）
+
+`best.ckpt` 是 PyTorch Lightning checkpoint，通常需要配合本项目的 `InkEraserModel` 代码使用，并提供 ResNet50 预训练权重文件（例如 `pretrained_weights/resnet50-0676ba61.pth`）。
+
+```python
+import torch
+from model import InkEraserModel
+
+model = InkEraserModel.load_from_checkpoint(
+    "best.ckpt",
+    weight="pretrained_weights/resnet50-0676ba61.pth",
+)
+model.eval()
+
+with torch.no_grad():
+    y = model(x)
+```
+
+## 训练超参数（来自 `config.json`）
+
+以下参数主要用于训练/复现，推理不必关心：
+
+```json
+{
+  "lr": 0.0001,
+  "weight_decay": 0.01,
+  "loss_w_charb": 0.78,
+  "loss_w_ssim": 0.16,
+  "loss_w_edge": 0.06,
+  "use_mask_loss": true,
+  "loss_mask_weight": 10.0,
+  "charbonnier_eps": 0.001
+}
+```
+
+## 许可证
+
+MIT

best.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:03a24c5cf35357a002b86a22250a2e165bf0e97a868bfb3492eb9ab0843798ec
+size 614277544

config.json

@@ -0,0 +1,24 @@
+{
+  "architectures": [
+    "InkEraserModel"
+  ],
+  "model_type": "unet_plus_plus",
+  "encoder_name": "resnet50",
+  "in_channels": 3,
+  "classes": 3,
+  "decoder_attention_type": "scse",
+  "activation": "sigmoid",
+  "framework": "pytorch-lightning",
+  "training_config": {
+    "lr": 0.0001,
+    "weight_decay": 0.01,
+    "loss_w_charb": 0.78,
+    "loss_w_ssim": 0.16,
+    "loss_w_edge": 0.06,
+    "use_mask_loss": true,
+    "loss_mask_weight": 10.0,
+    "charbonnier_eps": 0.001
+  },
+  "description": "Ink Erasure Model - Handwritten ink removal using U-Net++",
+  "license": "MIT"
+}

configuration.json

@@ -0,0 +1 @@
+{"framework":"Pytorch","task":"image-to-image"}

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:78acd1c02b274179ae975134eb827ed0bdb34d4e5890b7981f8dde08bb35eb7a
+size 204806056