Update README.md

README.md

@@ -37,6 +37,7 @@ are a common source of hallucinated OCR output.
   [olmOCR-mix-0225](https://huggingface.co/datasets/allenai/olmOCR-mix-0225). FUNSD is available
   for non-commercial, research and educational purposes only, and it was what kept v0
   non-commercial.
+- **CORD dataset added to the training data.** 
 
 The previous release, [`Inceptive/document-gate-keeper-v0`](https://huggingface.co/Inceptive/document-gate-keeper-v0),
 stays available unchanged. **v0.5 expects different preprocessing**, so do not reuse v0's code (see
@@ -50,7 +51,7 @@ stays available unchanged. **v0.5 expects different preprocessing**, so do not r
 | **Input** | `input_IMAGE`, `float32[N, 3, 640, 640]`, NCHW, RGB, normalized |
 | **Output** | `output_0`, `float32[N, 2]`, softmax probabilities, order `[bad, good]` |
 | **Runtime** | Any ONNX Runtime execution provider (CPU is fast enough for per-crop use) |
-| **Test F1** | 0.9951 on 1,016 held-out crops (see [Evaluation](#evaluation)) |
+| **Test F1** | 0.9857 on 1,396 held-out crops (see [Evaluation](#evaluation)) |
 
 ## Intended use
 
@@ -101,7 +102,7 @@ Feeding v0.5 with v0's `preprocess()` gives it inputs it was not trained on.
 `float32` tensor of shape `[N, 2]`, holding class probabilities in the order `["bad", "good"]`. The
 graph already ends with a `Softmax`, so do **not** apply another one.
 
-The decision rule we use is `P(good) < 0.8 → low quality`. Raise the threshold to skip more regions
+The decision rule we use is `P(good) < 0.6 → low quality`. Raise the threshold to skip more regions
 (fewer hallucinations, more missed text), or lower it to keep more.
 
 ## How to use
@@ -190,25 +191,25 @@ crop through `preprocess` above. Do not run the classifier on the detector's own
 
 ## Evaluation
 
-The model was evaluated on a held-out test split (10% of the dataset): 1,016 crops, 516 `bad` and
-500 `good`.
+The model was evaluated on a held-out test split (10% of the dataset): 1,396 crops, 688 `bad` and
+708 `good`.
 
 | Class | Precision | Recall | F1 | Error rate | Crops |
 |---|---|---|---|---|---|
-| `bad` | 0.9942 | 0.9961 | 0.9952 | 0.39% | 516 |
-| `good` | 0.9960 | 0.9940 | 0.9950 | 0.60% | 500 |
-| **Overall** | **0.9951** | **0.9951** | **0.9951** | **0.49%** | **1,016** |
+| `bad` | 0.9744 | 0.9971 | 0.9856 | 0.29% | 688 |
+| `good` | 0.9971 | 0.9746 | 0.9857 | 2.5% | 708 |
+| **Overall** | **0.9857** | **0.9857** | **0.9857** | **1.4%** | **1,396** |
 
 Confusion matrix (rows: true class, columns: predicted class):
 
 | | predicted `bad` | predicted `good` |
 |---|---|---|
-| **`bad`** | 514 | 2 |
-| **`good`** | 3 | 497 |
+| **`bad`** | 686 | 2 |
+| **`good`** | 18 | 690 |
 
-That is 5 errors out of 1,016 crops, an accuracy of 99.51%. The per-class error rate is
+That is 20 errors out of 1,396 crops, an accuracy of 98.57%. The per-class error rate is
 `1 − recall`. The overall precision and recall are micro-averaged, so they both equal the accuracy,
-and the overall F1 is the macro average of the two class F1 scores. The average test loss is 0.2346.
+and the overall F1 is the macro average of the two class F1 scores. The average test loss is 0.2581.
 
 > These crops come from the same synthetic pipeline as the training crops (see
 > [Training data](#training-data)). The numbers measure how well the model detects resampling and
@@ -238,7 +239,7 @@ print({p.key: p.value for p in onnx.load(model_path).metadata_props})
 
 ## Training data
 
-The dataset holds **10,142 crops**, split 90% / 10% into about 9,126 training and 1,016 test crops.
+The dataset holds **13,947 crops**, split 90% / 10% into about 12,522 training and 1,396 test crops.
 It is synthetic and was built from a single public source:
 
 - [**olmOCR-mix-0225**](https://huggingface.co/datasets/allenai/olmOCR-mix-0225) (Allen Institute
@@ -262,14 +263,14 @@ Unlike v0, no FUNSD data is used.
 | Setting | Value |
 |---|---|
 | Epochs | 10 |
-| Batch size | 32 |
+| Batch size | 16 |
 | Optimizer | AdamW, weight decay 1e-4 |
 | Learning rate | 5e-4, constant |
 | Loss | Cross-entropy with label smoothing 0.1 |
 | Input | 640 × 640 × 3: RGB, aspect-preserving resize (upscaling allowed), centered white padding, ImageNet normalization |
 
-The final training loss is 0.2409 (test: 0.2346). With label smoothing 0.1 and two classes, the
-lowest reachable loss is about 0.199, so both values are close to that floor despite the ~99.5%
+The final training loss is 0.2581 (test: 0.2589). With label smoothing 0.1 and two classes, the
+lowest reachable loss is about 0.199, so both values are close to that floor despite the ~98.5%
 accuracy.
 
 ## Limitations and biases