From 127ed2c4014a2b169c699407db28633019ec84ac Mon Sep 17 00:00:00 2001 From: hanwenbo Date: Mon, 21 Jul 2025 21:12:29 +0800 Subject: [PATCH] fix: add support for numpy integer types in JSON serialization MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit 修复 _format_data 函数未处理 numpy 整数类型（np.int32, np.int64）的问题， 防止 OCR 结果中包含 numpy 整数坐标时出现 JSON 序列化错误。 修复内容： - 添加对 np.int32 和 np.int64 类型的支持 - 将 numpy 整数转换为 Python 原生 int 类型 - 扩展 np.float64 支持以保持类型处理一致性 Fixes JSON serialization errors when OCR results contain numpy integer coordinates. 🤖 Generated with [Claude Code](https://claude.ai/code) Co-Authored- Claude --- paddlex/inference/common/result/mixin.py | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) diff --git a/paddlex/inference/common/result/mixin.py b/paddlex/inference/common/result/mixin.py index 70f8164365..4a58ebef75 100644 --- a/paddlex/inference/common/result/mixin.py +++ b/paddlex/inference/common/result/mixin.py @@ -79,8 +79,10 @@ def _format_data(obj): Returns: Any: The formatted object. """ - if isinstance(obj, np.float32): + if isinstance(obj, (np.float32, np.float64)): return float(obj) + elif isinstance(obj, (np.int32, np.int64)): + return int(obj) elif isinstance(obj, np.ndarray): return [_format_data(item) for item in obj.tolist()] elif isinstance(obj, pd.DataFrame):