CNWei/ddddocr-rs
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

refactor(error): 规范化分层错误类型并优化异常捕捉

Browse Source 
- 新增 tracing 记录异常，移除不必要的 Result
- 重构 错误处理架构
This commit is contained in:
CNWei
2026-07-20 20:18:38 +08:00
parent 913ff4d884
commit 3499e89bf1
11 changed files with 140 additions and 101 deletions
Download Patch File Download Diff File Expand all files Collapse all files

19
ddddocr-core/src/det/executor.rs
View File

@@ -1,7 +1,7 @@
use anyhow::{Context, Result};
use crate::error::{Result, TensorError};
use image::{imageops::FilterType, DynamicImage, GenericImageView};
use std::fmt;
use ndarray::{prelude::*, s, Array2, Array3, Array4, Axis};
use std::fmt;
// use tract_onnx::prelude::{Tensor};
@@ -51,7 +51,7 @@ impl<'a> Detector<'a> {
self.get_bbox(image)
}
/// 2. preproc: 纯 Rust 实现 (替代 OpenCV)
fn preproc(&self, image: &DynamicImage, input_size: (u32, u32)) -> Result<(Array4<f32>, f32)> {
fn preproc(&self, image: &DynamicImage, input_size: (u32, u32)) -> (Array4<f32>, f32) {
let (target_h, target_w) = input_size;
let (img_w, img_h) = image.dimensions();
@@ -89,7 +89,7 @@ impl<'a> Detector<'a> {
}
}
Ok((array, r))
(array, r)
}
/// 3. demo_postprocess (逻辑与 Python 一致)
@@ -252,7 +252,7 @@ impl<'a> Detector<'a> {
// let dynamic_img = image::load_from_memory(image_bytes).context("Failed to decode utils")?;
let (orig_w, orig_h) = dynamic_img.dimensions();
let (input_tensor, ratio) = self.preproc(dynamic_img, (416, 416))?;
let (input_tensor, ratio) = self.preproc(dynamic_img, (416, 416));
// tract 推理
// let outputs = self.session.session.run(tvec!(input_tensor.into()))?;
@@ -268,8 +268,13 @@ impl<'a> Detector<'a> {
let obj_conf = pred.slice(s![.., 4..5]);
let cls_conf = pred.slice(s![.., 5..]);
let obj_broadcast = obj_conf
.broadcast(cls_conf.dim())
.context("ndarray broadcasting failed for scores calculation")?;
.broadcast(cls_conf.dim()).ok_or_else(|| {
TensorError::DimensionMismatch {
expected: format!("可广播至 cls_conf 形状 {:?}", cls_conf.shape()),
actual: obj_conf.shape().to_vec(),
}
})?;
// .context("ndarray broadcasting failed for scores calculation")?;
let scores = &obj_broadcast * &cls_conf;
// let scores = &pred.slice(s![.., 4..5]) * &pred.slice(s![.., 5..]);

76
ddddocr-core/src/error.rs
View File

@@ -73,7 +73,8 @@ use thiserror::Error;
// }
/// 统一用我们自己的 DdddError 包装 Result
pub type Result<T> = std::result::Result<T, DdddError>;
// pub type Result<T> = std::result::Result<T, DdddError>;
pub type Result<T, E = DdddError> = std::result::Result<T, E>;
// =====================================================================
// 1. 顶层全局 Error 分流器 (去 anyhow 化，完全基于标准库/自定义类型)
// =====================================================================
@@ -83,27 +84,26 @@ pub enum DdddError {
// /// 系统文件、网络等原生 I/O 异常 (高优先级自动转换)
// #[error("系统网络或文件 I/O 异常: {0}")]
// Io(#[from] std::io::Error),
/// 图像预处理阶段发生异常
#[error("图像预处理失败: {0}")]
Preprocess(#[from] ImagePreprocessReason),
Preprocess(#[from] ImagePreprocessError),
/// 推理引擎与张量操作阶段发生异常
#[error("推理与模型输入/输出张量异常: {0}")]
Inference(#[from] TensorErrorReason),
Inference(#[from] TensorError),
/// 算法后处理解码阶段发生异常
#[error("后处理解码错误: {0}")]
Decode(#[from] DecodeReason),
Decode(#[from] DecodeError),
/// 框架内部不可恢复的逻辑断言错误（例如解析节点 Fact 失败）
#[error("内部严重逻辑错误: {0}")]
Internal(String),
/// 【流派一核心】接替 anyhow::Error 的用户自定义扩展错误
/// 【流派核心】接替 anyhow::Error 的用户自定义扩展错误
/// 承载任何第三方扩展、解密、特定预处理插件在执行时产生的自定义错误
#[error("用户自定义扩展错误: {0}")]
Other(Box<dyn std::error::Error + Send + Sync>),
Other(#[source]Box<dyn std::error::Error + Send + Sync>),
}
// =====================================================================
@@ -111,7 +111,7 @@ pub enum DdddError {
// =====================================================================
#[derive(Error, Debug)]
pub enum ImagePreprocessReason {
pub enum ImagePreprocessError {
// #[error("图片加载或解码失败: {0}")]
// ImageIo(#[from] image::ImageError),
// image_io
@@ -119,14 +119,16 @@ pub enum ImagePreprocessReason {
Ndarray(#[from] ndarray::ShapeError),
// image_io
#[error("图像矩阵维度不合规！预期: {expected}，实际图像形状: {actual:?}")]
InvalidImageDimensions {
InvalidDimensions {
expected: String,
actual: Vec<usize>,
},
// image_io
/// 从 ndarray 原始数据构建图像缓冲区时，缓冲区长度与分辨率/通道数不匹配
#[error("图像缓冲区长度不匹配！预期大小: {expected}，实际大小: {actual} (分辨率: {width}x{height}, 通道数: {channels})")]
#[error(
"图像缓冲区长度不匹配！预期大小: {expected}，实际大小: {actual} (分辨率: {width}x{height}, 通道数: {channels})"
)]
BufferLengthMismatch {
expected: usize,
actual: usize,
@@ -137,7 +139,18 @@ pub enum ImagePreprocessReason {
// image_io
#[error("不支持的图像通道数: {0} (仅支持单通道灰度L、3通道RGB、4通道RGBA)")]
UnsupportedChannels(usize),
// ================= 新增：针对 HSV 和 Preset 的强类型错误 =================
/// HSV 颜色区间非法 (例如 H > 180 或 lower > upper)
#[error("HSV 颜色区间参数非法: {0}")]
InvalidHsvRange (String ),
/// 不支持或未知的颜色预设名称
#[error("不支持的颜色预设名称: {0}")]
UnknownColorPreset(String),
/// 颜色过滤器/预处理规则配置非法导致失败
#[error("颜色过滤器配置无效或初始化失败: {0}")]
FilterConfigInvalid(String),
// #[error("Base64 解码失败: {0}")]
// Base64(#[from] base64::DecodeError),
//
@@ -153,14 +166,14 @@ pub enum ImagePreprocessReason {
// =====================================================================
#[derive(Error, Debug)]
pub enum TensorErrorReason {
pub enum TensorError {
/// 替换原有的 anyhow::Error，明确将 Tract/ONNX 引擎底层报错序列化为干净的 String
#[error("推理引擎内部发生异常: {0}")]
EngineError(String),
Engine(String),
/// 模型张量维度不匹配 (原有的顶层 DimensionMismatch 被优雅地归入本模块)
#[error("模型张量维度不匹配！预期: {expected}，实际 Tensor 形状: {actual:?}")]
TensorDimensionMismatch {
DimensionMismatch {
expected: String,
actual: Vec<usize>,
},
@@ -169,7 +182,7 @@ pub enum TensorErrorReason {
/// 直接包装 ndarray::ShapeError，保留强类型，完美支持 match
#[error("OCR Logits 矩阵变形失败: {0}")]
LogitsDimensionMismatch(#[from] ndarray::ShapeError),
/// 张量内存布局不是连续的
#[error("内存不连续，无法执行零拷贝操作")]
NonContiguousMemory,
@@ -184,10 +197,10 @@ pub enum TensorErrorReason {
// =====================================================================
#[derive(Error, Debug)]
pub enum DecodeReason {
pub enum DecodeError {
/// CTC 解码器解码过程中的逻辑报错
#[error("CTC 解码异常: {0}")]
CtcDecodeError(String),
Ctc(String),
}
// =====================================================================
@@ -223,7 +236,7 @@ impl DdddError {
pub fn is_invalid_dimensions(&self) -> bool {
matches!(
self,
DdddError::Preprocess(ImagePreprocessReason::InvalidImageDimensions { .. })
DdddError::Preprocess(ImagePreprocessError::InvalidDimensions { .. })
)
}
@@ -231,22 +244,21 @@ impl DdddError {
pub fn is_unsupported_channels(&self) -> bool {
matches!(
self,
DdddError::Preprocess(ImagePreprocessReason::UnsupportedChannels(_))
DdddError::Preprocess(ImagePreprocessError::UnsupportedChannels(_))
)
}
/// 提取出底层最原始的那个错误（无论是 IO、预处理、推理、还是第三方扩展错误）
/// 方便外层统一打印更深层的 `source` 链条
pub fn source_error(&self) -> Option<&(dyn std::error::Error + 'static)> {
use std::error::Error;
match self {
// DdddError::Io(err) => Some(err),
DdddError::Preprocess(err) => Some(err),
DdddError::Inference(err) => Some(err),
DdddError::Decode(err) => Some(err),
DdddError::Other(err) => Some(err.as_ref()),
DdddError::Internal(_) => None, // Internal 内部目前只有 String，没有底层的 Error source
}
}
// 提取出底层最原始的那个错误（无论是 IO、预处理、推理、还是第三方扩展错误）
// 方便外层统一打印更深层的 `source` 链条
// pub fn source_error(&self) -> Option<&(dyn std::error::Error + 'static)> {
// use std::error::Error;
// match self {
// // DdddError::Io(err) => Some(err),
// DdddError::Preprocess(err) => Some(err),
// DdddError::Inference(err) => Some(err),
// DdddError::Decode(err) => Some(err),
// DdddError::Other(err) => Some(err.as_ref()),
// DdddError::Internal(_) => None, // Internal 内部目前只有 String，没有底层的 Error source
// }
// }
}

44
ddddocr-core/src/ocr/color_filter.rs
View File

@@ -1,5 +1,5 @@
use crate::error::{ImagePreprocessError, Result};
use crate::utils::image_processor::rgb_to_opencv_hsv;
use anyhow::anyhow;
use image::{DynamicImage, ImageBuffer, Rgb};
use std::str::FromStr;
@@ -18,12 +18,13 @@ fn is_pixel_matched(ranges: &[HsvRange], h: u8, s: u8, v: u8) -> bool {
pub fn apply_to_image(
image: &DynamicImage,
hsv_ranges: &[HsvRange],
) -> anyhow::Result<DynamicImage> {
) -> Result<DynamicImage, ImagePreprocessError> {
// 1. 统一转换为连续内存的 RGB8 缓冲区 (对应 Python 的 Image 到 RGB/BGR 数组转换)
let rgb_img = image.to_rgb8();
let (width, height) = rgb_img.dimensions();
let mut raw_pixels = rgb_img.into_raw();
let actual_len = raw_pixels.len();
let expected_len = (width as usize) * (height as usize) * 3;
// 2. 密集计算核心：原地流式迭代修改
// 每次取出 3 个 u8 字节，分别代表 [R, G, B]，无多余掩膜矩阵内存分配
for chunk in raw_pixels.chunks_exact_mut(3) {
@@ -45,8 +46,14 @@ pub fn apply_to_image(
// 3. 将扁平字节数组重新打包回 DynamicImage 容器
let filtered_buffer = ImageBuffer::<Rgb<u8>, Vec<u8>>::from_raw(width, height, raw_pixels)
.ok_or_else(|| anyhow!("图像缓冲重新组装失败，维度与数据大小不匹配"))?;
// .ok_or_else(|| anyhow!("图像缓冲重新组装失败，维度与数据大小不匹配"))?;
.ok_or_else(|| ImagePreprocessError::BufferLengthMismatch {
expected: expected_len,
actual: actual_len,
width,
height,
channels: 3,
})?;
Ok(DynamicImage::ImageRgb8(filtered_buffer))
}
@@ -64,16 +71,22 @@ impl HsvRange {
impl HsvRange {
/// 验证当前 HSV 范围是否合法
/// 对应 Python 逻辑：H 在 0-180，S/V 在 0-255，且下界 <= 上界
pub fn validate(&self) -> Result<(), String> {
pub fn validate(&self) -> Result<(), ImagePreprocessError> {
// 1. 校验 H 通道边界 (OpenCV 中 H 范围是 0-180)
if self.lower.0 > 180 || self.upper.0 > 180 {
return Err("H通道值必须在 0-180 范围内".to_string());
// return Err("H通道值必须在 0-180 范围内".to_string());
return Err(ImagePreprocessError::InvalidHsvRange(
"H通道值必须在 0-180 范围内".to_string(),
));
}
// 2. 校验下界不能大于上界
if self.lower.0 > self.upper.0 || self.lower.1 > self.upper.1 || self.lower.2 > self.upper.2
{
return Err("HSV范围下界不能大于上界".to_string());
// return Err("HSV范围下界不能大于上界".to_string());
return Err(ImagePreprocessError::InvalidHsvRange(
"HSV范围下界不能大于上界".to_string(),
));
}
Ok(())
@@ -150,7 +163,7 @@ impl ColorPreset {
}
}
/// 校验逻辑：在这里实现完美的“责任分离”
pub fn validate(&self) -> Result<(), String> {
pub fn validate(&self) -> Result<(), ImagePreprocessError> {
match self {
// 1. 快捷变体：完全绕过，根本不校验，0 运行时开销放行！
ColorPreset::Custom(ranges) => {
@@ -166,7 +179,7 @@ impl ColorPreset {
}
impl FromStr for ColorPreset {
type Err = String;
type Err = ImagePreprocessError;
fn from_str(s: &str) -> Result<Self, Self::Err> {
match s.to_lowercase().as_str() {
"red" => Ok(ColorPreset::Red),
@@ -179,7 +192,8 @@ impl FromStr for ColorPreset {
"black" => Ok(ColorPreset::Black),
"white" => Ok(ColorPreset::White),
"gray" => Ok(ColorPreset::Gray),
_ => Err(format!("不支持的颜色预设: {}", s)),
// _ => Err(format!("不支持的颜色预设: {}", s)),
_ => Err(ImagePreprocessError::UnknownColorPreset(s.to_string())),
}
}
}
@@ -200,12 +214,12 @@ pub trait ColorFilter {
fn estimated_count(&self) -> usize;
/// 将自身的有效约束平铺追加到统一目标容器中
/// 验证当前过滤器是否合法，默认直接放行（Ok(())）
fn validate_self(&self) -> Result<(), String> {
fn validate_self(&self) -> Result<(), ImagePreprocessError> {
Ok(())
}
/// 【新扩展的架构方法】将自身安全的合并到已有的普通容器中，并完成去重和排序
/// 完美的责任分离：Builder 不再需要关心怎么分配内存、怎么排序去重
fn collect_to_vec(&self) -> Result<Option<Vec<HsvRange>>, String> {
fn collect_to_vec(&self) -> Result<Option<Vec<HsvRange>>, ImagePreprocessError> {
// 1. 触发自检
self.validate_self()?;
@@ -237,7 +251,7 @@ impl ColorFilter for ColorPreset {
// 直接获取切片长度
self.matches().len()
}
fn validate_self(&self) -> Result<(), String> {
fn validate_self(&self) -> Result<(), ImagePreprocessError> {
// 直接调用我们在第一步中为 ColorPreset 实现的精细化分流校验
// 快捷变体在这里会直接返回 Ok(()), 只有 Custom 才会去真正校验
self.validate()
@@ -261,7 +275,7 @@ impl<'a> ColorFilter for MultiOrColorRestrict<'a> {
self.filters.iter().map(|f| f.estimated_count()).sum()
}
fn validate_self(&self) -> Result<(), String> {
fn validate_self(&self) -> Result<(), ImagePreprocessError> {
// 递归政审：只要其中一个子过滤器校验失败（比如某个 Custom 变体非法），立刻熔断
for f in &self.filters {
f.validate_self()?;

41
ddddocr-core/src/ocr/executor.rs
View File

@@ -4,7 +4,6 @@ use crate::ocr::color_filter::{HsvRange, apply_to_image};
// use ddddocr_tract::session::{ModelOutput, OcrSession};
use crate::utils::image_convert::png_rgba_white_preprocess;
use crate::utils::image_processor::{convert_to_grayscale, resize_image};
use anyhow::Result;
use image::DynamicImage;
use serde::Serialize;
use std::borrow::Cow;
@@ -17,7 +16,9 @@ use ndarray::ArrayView2;
// Indices(ndarray::Array1<i64>), // 拥有完整所有权的 1维数组，可任意传递和返回
// Logits(ndarray::Array2<f32>), // 拥有完整所有权的 2维矩阵，可任意传递和返回
// }
use crate::error::{ImagePreprocessError, Result, TensorError};
use crate::{OcrEngine, OcrOutput};
use tracing::{ warn};
#[derive(Debug, Clone, Serialize)]
pub enum OcrResult {
/// 纯文本分支（对应 probability = false）
@@ -107,7 +108,7 @@ pub struct Ocr<'a> {
pub(crate) png_fix: bool,
pub(crate) probability: bool,
/// 颜色过滤：保留的颜色列表
pub(crate) final_color_ranges: Result<Option<Vec<HsvRange>>, String>,
pub(crate) final_color_ranges: Result<Option<Vec<HsvRange>>, ImagePreprocessError>,
/// 字符集范围
pub(crate) final_charset_indices: Option<Vec<usize>>,
@@ -127,7 +128,7 @@ impl<'a> Ocr<'a> {
}
}
impl<'a> Ocr<'a> {
pub fn predict(&self, image: &DynamicImage) -> anyhow::Result<OcrResult> {
pub fn predict(&self, image: &DynamicImage) -> Result<OcrResult> {
println!("当前颜色过滤器状态: {:?}", self.final_color_ranges);
// =====================================================================
@@ -137,10 +138,13 @@ impl<'a> Ocr<'a> {
// =====================================================================
let img_cow = match &self.final_color_ranges {
Err(err_msg) => {
return Err(anyhow::anyhow!(
"颜色过滤器初始化失败，全链路短路: {}",
err_msg
));
// return Err(anyhow::anyhow!(
// "颜色过滤器初始化失败，全链路短路: {}",
// err_msg
// ));
return Err(ImagePreprocessError::FilterConfigInvalid(
err_msg.to_string(),
))?;
}
Ok(None) => {
// 核心优化点：直接借用原图，不发生任何克隆
@@ -168,12 +172,12 @@ impl<'a> Ocr<'a> {
// let raw_indices = self.ocr.extract_indices_from_tensor(&raw_tensor)?;
// // 步骤 2: 将索引切片 `&[i64]` 传给解码器进行 CTC 去重和字符映射
// let final_text = self.ctc_decode_to_string(&raw_indices);
let ocr_output = self.process_model_output(raw_tensor);
ocr_output
let ocr_output = self.process_model_output(raw_tensor)?;
Ok(ocr_output)
}
/// 对应 Python 的 _preprocess_image
/// 负责：透明背景修复 -> 灰度化 -> 按比例 Resize -> 归一化 -> 4维张量转换
fn preprocess_image(&self, img: &DynamicImage) -> anyhow::Result<ndarray::Array4<f32>> {
fn preprocess_image(&self, img: &DynamicImage) -> Result<ndarray::Array4<f32>,ImagePreprocessError> {
// 1. 获取模型元数据配置
let meta = self.session.metadata();
let norm = &meta.normalization; // 获取归一化器
@@ -239,7 +243,12 @@ impl<'a> Ocr<'a> {
array
}
_ => return Err(anyhow::anyhow!("不支持的通道数配置: {}", meta.channel)),
// _ => return Err(anyhow::anyhow!("不支持的通道数配置: {}", meta.channel)),
_ => {
return Err(ImagePreprocessError::UnsupportedChannels(
meta.channel as usize,
));
}
};
Ok(array4)
// Ok(tensor)
@@ -265,13 +274,14 @@ impl<'a> Ocr<'a> {
}
// 这段代码未来直接放入 ddddocr-core
fn process_model_output(&self, output: OcrOutput) -> anyhow::Result<OcrResult> {
fn process_model_output(&self, output: OcrOutput) -> Result<OcrResult,TensorError> {
match output {
OcrOutput::Indices(array1) => {
// 对应你原来的 process_i64_tensor
let slice = array1
.as_slice()
.ok_or_else(|| anyhow::anyhow!("内存不连续，无法执行零拷贝解码"))?;
// .ok_or_else(|| anyhow::anyhow!("内存不连续，无法执行零拷贝解码"))?;
.ok_or_else(|| TensorError::NonContiguousMemory)?;
let final_text = self.ctc_decode_to_string(slice);
if self.probability {
@@ -528,8 +538,9 @@ impl<'a> Ocr<'a> {
// 5. 字符映射
if let Some(char_str) = tokens.get(u_idx) {
res.push_str(char_str);
} else {
eprintln!("警告: 预测索引 {} 超出字符集范围", u_idx);
}
else {
warn!("警告: 预测索引 {} 超出字符集范围", u_idx);
}
}
res

2
ddddocr-core/src/ocr/metadata.rs
View File

@@ -1,4 +1,4 @@
use anyhow::{anyhow, Result};
use crate::error::{ Result};
use serde::Deserialize;
use std::borrow::Cow;
use std::collections::HashMap;

34
ddddocr-core/src/utils/image_convert.rs
View File

@@ -1,4 +1,4 @@
use crate::error::{DdddError, ImagePreprocessReason, Result};
use crate::error::{DdddError, ImagePreprocessError, Result};
use image::{DynamicImage, GenericImageView, ImageBuffer, Luma, Rgb, Rgba};
use ndarray::{Array3, ArrayViewD};
@@ -35,19 +35,16 @@ pub fn ndarray_to_hwc_image(array: ArrayViewD<u8>) -> Result<DynamicImage> {
// 对应 Python: array.shape[2] == 4 (RGBA H, W, 4)
4 => ColorMode::RGBA,
_ => {
return Err(DdddError::Preprocess(
ImagePreprocessReason::UnsupportedChannels(c),
));
return Err(ImagePreprocessError::UnsupportedChannels(c))?;
}
}
}
_ => {
return Err(DdddError::Preprocess(
ImagePreprocessReason::InvalidImageDimensions {
expected: "2D (H,W) 或 3D (H,W,C)".to_string(),
actual: shape.to_vec(),
},
));
return Err(ImagePreprocessError::InvalidDimensions {
expected: "2D (H,W) 或 3D (H,W,C)".to_string(),
actual: shape.to_vec(),
}
.into());
}
};
from_ndarray(array, color_mode)
@@ -113,7 +110,7 @@ pub fn image_to_ndarray(image: &DynamicImage, mode: ColorMode) -> Result<Array3<
};
let array = Array3::from_shape_vec((height as usize, width as usize, channels), raw)
.map_err(ImagePreprocessReason::from)?;
.map_err(ImagePreprocessError::from)?;
Ok(array)
}
/// 将 array 数组转换为 DynamicImage
@@ -122,12 +119,10 @@ pub fn ndarray_to_image(array: ArrayViewD<u8>, mode: ColorMode) -> Result<Dynami
// 基础边界检查：至少要有 H 和 W 两个维度
if shape.len() < 2 {
return Err(DdddError::Preprocess(
ImagePreprocessReason::InvalidImageDimensions {
expected: "至少为 2D array [H, W]".to_string(),
actual: shape.to_vec(),
},
));
return Err(ImagePreprocessError::InvalidDimensions {
expected: "至少为 2D array [H, W]".to_string(),
actual: shape.to_vec(),
})?;
}
from_ndarray(array, mode)
}
@@ -156,13 +151,14 @@ fn from_ndarray(array: ArrayViewD<u8>, mode: ColorMode) -> Result<DynamicImage>
// 构造通用错误闭包，避免 match 分支中重复编写冗长的错误对象
let make_err = || {
DdddError::Preprocess(ImagePreprocessReason::BufferLengthMismatch {
ImagePreprocessError::BufferLengthMismatch {
expected: expected_len,
actual: raw_len,
width,
height,
channels,
})
}
.into()
};
// 2. 重新解释内存并构建 ImageBuffer

4
ddddocr-core/src/utils/image_helper.rs
View File

@@ -1,6 +1,6 @@
use crate::error::{DdddError, ImagePreprocessReason, Result};
use crate::error::{DdddError, Result};
use crate::utils::image_convert::ndarray_to_hwc_image;
use base64::{Engine as _, engine::general_purpose};
use base64::{engine::general_purpose, Engine as _};
use image::DynamicImage;
use ndarray::ArrayViewD;
use std::fmt;

14
ddddocr-core/src/utils/tensor_transform.rs
View File

@@ -1,8 +1,8 @@
use crate::OcrOutput;
use crate::error::{DdddError, Result, TensorErrorReason};
use crate::error::{DdddError, Result, TensorError};
use ndarray::s;
/// 核心层复用资产：将异构的动态维度矩阵转化为标准 OCR 2D Logits 矩阵
pub fn normalize_ocr_logits(array: ndarray::ArrayViewD<f32>, shape: &[usize]) -> Result<OcrOutput> {
pub fn normalize_ocr_logits(array: ndarray::ArrayViewD<f32>, shape: &[usize]) -> Result<OcrOutput,TensorError> {
let (steps, classes, data_dyn_view) = match shape.len() {
3 => {
if shape[1] == 1 {
@@ -24,12 +24,12 @@ pub fn normalize_ocr_logits(array: ndarray::ArrayViewD<f32>, shape: &[usize]) ->
// 我们把它虚构成一个 [1, Classes] 的 2D 矩阵来复用后面的 argmax 逻辑
1 => (1, shape[0], array),
_ => {
return Err(DdddError::Inference(
TensorErrorReason::TensorDimensionMismatch {
return Err(
TensorError::DimensionMismatch {
expected: "1D, 2D, or 3D OCR Logits".to_string(),
actual: shape.to_vec(),
},
));
);
}
};
@@ -39,10 +39,10 @@ pub fn normalize_ocr_logits(array: ndarray::ArrayViewD<f32>, shape: &[usize]) ->
.map_err(|shape_err| {
// 如果是因为切片导致不连续且无法进行零拷贝变换，抛出 NonContiguousMemory
if !data_dyn_view.is_standard_layout() {
DdddError::Inference(TensorErrorReason::NonContiguousMemory)
TensorError::NonContiguousMemory
} else {
// 否则，说明是纯粹的数据元素数量不对（Shape 不匹配），抛出专属的强类型错误
DdddError::Inference(TensorErrorReason::LogitsDimensionMismatch(shape_err))
TensorError::LogitsDimensionMismatch(shape_err)
}
})?
.to_owned();