Re-export CRAFT as single .onnx file (fixes GPU acceleration)

README.md

@@ -21,8 +21,8 @@ Converted for use with [inference4j](https://github.com/inference4j/inference4j)
 ## Usage with inference4j
 
 ```java
-try (Craft craft = Craft.fromPretrained("models/craft-mlt-25k")) {
-    List<TextRegion> regions = craft.detect(Path.of("document.jpg"));
+try (CraftTextDetector detector = CraftTextDetector.builder().build()) {
+    List<TextRegion> regions = detector.detect(Path.of("document.jpg"));
     for (TextRegion r : regions) {
         System.out.printf("Text at [%.0f, %.0f, %.0f, %.0f] (confidence=%.2f)%n",
             r.box().x1(), r.box().y1(), r.box().x2(), r.box().y2(),
@@ -62,21 +62,6 @@ try (Craft craft = Craft.fromPretrained("models/craft-mlt-25k")) {
 4. For each component: compute mean region score, filter by `text_threshold` (default 0.7)
 5. Extract axis-aligned bounding box, scale back to original image coordinates
 
-## Conversion
-
-This model was converted from PyTorch to ONNX using [`craft_exporter.py`](craft_exporter.py) included in this repo. To reproduce:
-
-```bash
-# Download original PyTorch weights (~79 MB)
-gdown 1Jk4eGD7crsqCCg9C9VjCLkMN3ze8kutZ -O craft_mlt_25k.pth
-
-# Install dependencies
-pip install torch torchvision onnx onnxruntime
-
-# Export to ONNX
-python craft_exporter.py
-```
-
 ## Original Paper
 
 > Baek, Y., Lee, B., Han, D., Yun, S., & Lee, H. (2019).

model.onnx

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b0773559ece3671df618d660e68782406ccfb830c18281664b453e5fe97c95bb
-size 132927
+oid sha256:28c05f8b5e97f3c5765d00bca556ec2397c86c954d4ac5ef73aa10043db16388
+size 83169398