FastDeploy项目简介，使用其进行（图像分类、目标检测、语义分割、文本检测|orc部署）

万里鹏程转瞬至2024-01-29 9:18

FastDeploy是一款全场景、易用灵活、极致高效的AI推理部署工具， 支持云边端部署。提供超过 🔥160+ Text，Vision， Speech和跨模态模型📦开箱即用的部署体验，并实现🔚端到端的推理性能优化。包括 物体检测、字符识别（OCR）、人脸、人像扣图、多目标跟踪系统、NLP、Stable Diffusion文图生成、TTS 等几十种任务场景，满足开发者多场景、多硬件、多平台的产业部署需求。

1、FastDeploy支持的环境

1.1 芯片支持

FastDeploy支持各个常见的硬件平台和各种国产化的芯片。例如华为的ascend
X86_64 CPU
NVDIA GPU
飞腾 CPU
昆仑芯 XPU
华为昇腾 NPU
Graphcore IPU
算能
Intel 显卡
Jetson
ARM CPU
RK3588等
RV1126等
晶晨
恩智浦

1.2 编程语言支持

FastDeploy支持python、c++、java等开发语言，同时还支持web与小程序部署。

1.3 操作系统支持

FastDeploy支持linux、windows、mac、安卓部署。

1.4 推理后端支持

FastDeploy对各种推理后端接口均做了封装，可以以一种代码风格同时调用onnx、paddle infer、tensorrt、openvino等推理框架。

2、安装与使用

本博文对应的代码和模型资源可以进行付费下载，也可以参考博文资料轻松实现复现。

2.1 安装命令

🔸 前置依赖
  • CUDA >= 11.2、cuDNN >= 8.0、Python >= 3.6
  • OS: Linux x86_64/macOS/Windows 10
🔸 安装GPU版本
bash 复制代码 
pip install numpy opencv-python fastdeploy-gpu-python -f https://www.paddlepaddle.org.cn/whl/fastdeploy.html
🔸 Conda安装(推荐✨)
bash 复制代码 
conda config --add channels conda-forge && conda install cudatoolkit=11.2 cudnn=8.2
🔸 安装CPU版本
bash 复制代码 
pip install numpy opencv-python fastdeploy-python -f https://www.paddlepaddle.org.cn/whl/fastdeploy.html

最终安装成功的输出如下所示

2.2 部署图像分类模型

下载paddle官方的预训练模型，https://bj.bcebos.com/paddlehub/fastdeploy/ResNet50_vd_infer.tgz，也可以使用自己基于paddleclas训练的模型

将以下图片保存为000000014439.jpg

预测代码如下所示，使用onnxruntime-gpu进行推理

py 复制代码 
# GPU/TensorRT部署参考 examples/vision/detection/paddledetection/python
import cv2
import fastdeploy as fd
import fastdeploy.vision as vision
import ipdb

option = fd.RuntimeOption()
#option.use_trt_backend()
option.use_gpu(device_id=0) #使用0号gpu进行预测
option.use_ort_backend() #使用onnxruntime作为后端

model_file="ResNet50_vd_infer/inference.pdmodel"
params_file="ResNet50_vd_infer/inference.pdiparams"
config_file="ResNet50_vd_infer/inference_cls.yaml"
model = fd.vision.classification.PaddleClasModel(
    model_file, params_file, config_file, runtime_option=option)
im = cv2.imread("000000014439.jpg")
result = model.predict(im, topk=5)
print(type(result),dir(result))
print(result)

程序运行后输出如下所示：

py 复制代码 
[INFO] fastdeploy/vision/common/processors/transform.cc(93)::fastdeploy::vision::FuseNormalizeHWC2CHW                                                                  Normalize and HWC2CHW are fused to NormalizeAndPermute  in preprocessing pipeline.
[INFO] fastdeploy/vision/common/processors/transform.cc(159)::fastdeploy::vision::FuseNormalizeColorConvert                                                            BGR2RGB and NormalizeAndPermute are fused to NormalizeAndPermute with swap_rb=1
[INFO] fastdeploy/runtime/runtime.cc(300)::fastdeploy::Runtime::CreateOrtBackend                                                                                       Runtime initialized with Backend::ORT in Device::GPU.
<class 'fastdeploy.libs.fastdeploy_main.vision.ClassifyResult'> ['__class__', '__delattr__', '__dir__', '__doc__', '__eq__', '__format__', '__ge__', '__getattribute__', '__getstate__', '__gt__', '__hash__', '__init__', '__init_subclass__', '__le__', '__lt__', '__module__', '__ne__', '__new__', '__reduce__', '__reduce_ex__', '__repr__', '__setattr__', '__setstate__', '__sizeof__', '__str__', '__subclasshook__', 'feature', 'label_ids', 'scores']
ClassifyResult(
label_ids: 701, 879, 417, 645, 723, 
scores: 0.424514, 0.084369, 0.058738, 0.016847, 0.013349, 
)

从以上输出中可以看到，预测结果又'feature', 'label_ids', 'scores'三个属性。

2.2 部署目标检测模型

下载paddle官方的预训练模型，也可以使用自己导出的paddle infer模型。
https://bj.bcebos.com/paddlehub/fastdeploy/ppyoloe_crn_l_300e_coco.tgz

将以下图片保存为000000014439.jpg

预测代码如下所示

py 复制代码 
# GPU/TensorRT部署参考 examples/vision/detection/paddledetection/python
import cv2
import fastdeploy.vision as vision

model = vision.detection.PPYOLOE("ppyoloe_crn_l_300e_coco/model.pdmodel",
                                 "ppyoloe_crn_l_300e_coco/model.pdiparams",
                                 "ppyoloe_crn_l_300e_coco/infer_cfg.yml")
im = cv2.imread("000000014439.jpg")
result = model.predict(im)
print(result)

vis_im = vision.vis_detection(im, result, score_threshold=0.5)
cv2.imshow("vis_image.jpg", vis_im)
cv2.waitKey()
#cv2.imwrite("vis_image.jpg", vis_im)

程序运行时的输出如下所示，可见默认是使用OPENVINO cpu进行推理。其返回结果为DetectionResult对象，其一共有'boxes', 'contain_masks', 'label_ids', 'masks', 'rotated_boxes', 'scores'等属性。其中boxes是xmin, ymin, xmax, ymax,格式的数据，其默认只返回nms操作后的top 300个结果，后续需要自己根据scores进行结果过滤。

txt 复制代码 
[INFO] fastdeploy/vision/common/processors/transform.cc(45)::fastdeploy::vision::FuseNormalizeCast      Normalize and Cast are fused to Normalize in preprocessing pipeline.
[INFO] fastdeploy/vision/common/processors/transform.cc(93)::fastdeploy::vision::FuseNormalizeHWC2CHW   Normalize and HWC2CHW are fused to NormalizeAndPermute  in preprocessing pipeline.
[INFO] fastdeploy/vision/common/processors/transform.cc(159)::fastdeploy::vision::FuseNormalizeColorConvert     BGR2RGB and NormalizeAndPermute are fused to NormalizeAndPermute with swap_rb=1
[INFO] fastdeploy/runtime/backends/openvino/ov_backend.cc(218)::fastdeploy::OpenVINOBackend::InitFromPaddle     number of streams:1.
[INFO] fastdeploy/runtime/backends/openvino/ov_backend.cc(228)::fastdeploy::OpenVINOBackend::InitFromPaddle     affinity:YES.
[INFO] fastdeploy/runtime/backends/openvino/ov_backend.cc(240)::fastdeploy::OpenVINOBackend::InitFromPaddle     Compile OpenVINO model on device_name:CPU.
[INFO] fastdeploy/runtime/runtime.cc(286)::fastdeploy::Runtime::CreateOpenVINOBackend   Runtime initialized with Backend::OPENVINO in Device::CPU.
DetectionResult: [xmin, ymin, xmax, ymax, score, label_id]
415.047180,89.311569, 506.009613, 283.863098, 0.950423, 0
163.665710,81.914932, 198.585342, 166.760895, 0.896433, 0
581.788635,113.027618, 612.623474, 198.521713, 0.842596, 0
267.217224,89.777306, 298.796051, 169.361526, 0.837951, 0
104.465584,45.482422, 127.688850, 93.533867, 0.773348, 0
348.902130,44.059338, 367.541687, 98.403542, 0.767127, 0
363.889740,58.385532, 381.397522, 114.652367, 0.756467, 0
504.843811,114.531601, 612.280945, 271.292572, 0.714129, 0

2.4 部署语义分割模型

下载paddle发布的预训练模型SegFormer_B0，也可以使用paddleseg训练后导出模型。

https://bj.bcebos.com/paddlehub/fastdeploy/ppyoloe_crn_l_300e_coco.tgz

将以下图片保存为000000014439.jpg

预测代码如下所示

py 复制代码 
# GPU/TensorRT部署参考 examples/vision/detection/paddledetection/python
import cv2
import fastdeploy as fd
import fastdeploy.vision as vision
import numpy as np
import ipdb
def get_color_map_list(num_classes):
    """
    Args:
        num_classes (int): number of class
    Returns:
        color_map (list): RGB color list
    """
    color_map = num_classes * [0, 0, 0]
    for i in range(0, num_classes):
        j = 0
        lab = i
        while lab:
            color_map[i * 3] |= (((lab >> 0) & 1) << (7 - j))
            color_map[i * 3 + 1] |= (((lab >> 1) & 1) << (7 - j))
            color_map[i * 3 + 2] |= (((lab >> 2) & 1) << (7 - j))
            j += 1
            lab >>= 3
    color_map = [color_map[i:i + 3] for i in range(0, len(color_map), 3)]
    return color_map

color_map=get_color_map_list(80) #80个类别
color_map=np.array(color_map).astype(np.uint8)

option = fd.RuntimeOption()
#option.use_kunlunxin() #使用昆仑芯片进行部署
#option.use_lite_backend()
option.use_paddle_infer_backend()

model_file="SegFormer_B0-cityscapes-with-argmax/model.pdmodel"
params_file="SegFormer_B0-cityscapes-with-argmax/model.pdiparams"
config_file="SegFormer_B0-cityscapes-with-argmax/deploy.yaml"
model = fd.vision.segmentation.PaddleSegModel(
    model_file, params_file, config_file, runtime_option=option)
im = cv2.imread("000000014439.jpg")
result = model.predict(im)
print(type(result),dir(result))
print(result)
#ipdb.set_trace()
res=np.array(result.label_map,np.uint8)
res=res.reshape(result.shape)

res=color_map[res] #为结果上色
cv2.imshow("label_map",res)
cv2.waitKey()

代码运行时输出如下所示：

py 复制代码 
[INFO] fastdeploy/vision/common/processors/transform.cc(93)::fastdeploy::vision::FuseNormalizeHWC2CHW                                                                  Normalize and HWC2CHW are fused to NormalizeAndPermute  in preprocessing pipeline.
[INFO] fastdeploy/vision/common/processors/transform.cc(159)::fastdeploy::vision::FuseNormalizeColorConvert                                                            BGR2RGB and NormalizeAndPermute are fused to NormalizeAndPermute with swap_rb=1
WARNING: Logging before InitGoogleLogging() is written to STDERR
W0125 21:32:56.195503  5356 analysis_config.cc:971] It is detected that mkldnn and memory_optimize_pass are enabled at the same time, but they are not supported yet. Currently, memory_optimize_pass is explicitly disabled
[INFO] fastdeploy/runtime/runtime.cc(273)::fastdeploy::Runtime::CreatePaddleBackend                                                                                    Runtime initialized with Backend::PDINFER in Device::CPU.
<class 'fastdeploy.libs.fastdeploy_main.vision.SegmentationResult'> ['__class__', '__delattr__', '__dir__', '__doc__', '__eq__', '__format__', '__ge__', '__getattribute__', '__getstate__', '__gt__', '__hash__', '__init__', '__init_subclass__', '__le__', '__lt__', '__module__', '__ne__', '__new__', '__reduce__', '__reduce_ex__', '__repr__', '__setattr__', '__setstate__', '__sizeof__', '__str__', '__subclasshook__', 'contain_score_map', 'label_map', 'score_map', 'shape']
SegmentationResult Image masks 10 rows x 10 cols:
[4, 4, 4, 4, 4, 2, 2, 2, 2, 2, .....]
[2, 2, 2, 2, 2, 2, 2, 2, 2, 2, .....]
[2, 2, 2, 2, 2, 2, 2, 2, 2, 2, .....]
[2, 2, 2, 2, 2, 2, 2, 2, 2, 2, .....]
[2, 2, 2, 2, 2, 2, 2, 2, 2, 2, .....]
[2, 2, 2, 2, 2, 2, 2, 4, 4, 4, .....]
[4, 4, 4, 4, 4, 4, 4, 4, 4, 4, .....]
[4, 4, 2, 2, 2, 2, 2, 2, 2, 2, .....]
[2, 2, 2, 2, 2, 2, 2, 2, 2, 2, .....]
[2, 2, 2, 2, 2, 2, 2, 2, 2, 2, .....]
...........
result shape is: [404 640]

运行效果如下所示

2.5 部署文本识别模型

进行全流程文本识别

首先下载以下模型，可以手动打开http链接进行下载。

下载PP-OCRv3文字检测模型

wget https://paddleocr.bj.bcebos.com/PP-OCRv3/chinese/ch_PP-OCRv3_det_infer.tar
tar -xvf ch_PP-OCRv3_det_infer.tar

下载文字方向分类器模型

wget https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_cls_infer.tar
tar -xvf ch_ppocr_mobile_v2.0_cls_infer.tar

下载PP-OCRv3文字识别模型

wget https://paddleocr.bj.bcebos.com/PP-OCRv3/chinese/ch_PP-OCRv3_rec_infer.tar
tar -xvf ch_PP-OCRv3_rec_infer.tar

准备label文件

下载ppocr_keys_v1.txt，保存到代码目录下。

准备测试文件

将以下图片保存为12.jpg

测试代码如下所示;

py 复制代码 
# Copyright (c) 2022 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#    http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

import fastdeploy as fd
import cv2
import os

def parse_arguments():
    import argparse
    import ast
    parser = argparse.ArgumentParser()
    parser.add_argument(
        "--det_model", default="ch_PP-OCRv3_det_infer", help="Path of Detection model of PPOCR.")
    parser.add_argument(
        "--cls_model",default="ch_ppocr_mobile_v2.0_cls_infer",
        help="Path of Classification model of PPOCR.")
    parser.add_argument(
        "--rec_model",default="ch_PP-OCRv3_rec_infer",
        help="Path of Recognization model of PPOCR.")
    parser.add_argument(
        "--rec_label_file",
        default="ppocr_keys_v1.txt",
        help="Path of Recognization model of PPOCR.")
    parser.add_argument(
        "--image", type=str, default="12.jpg", help="Path of test image file.")
    parser.add_argument(
        "--device",
        type=str,
        default='cpu',
        help="Type of inference device, support 'cpu' or 'gpu'.")
    parser.add_argument(
        "--device_id",
        type=int,
        default=0,
        help="Define which GPU card used to run model.")
    parser.add_argument(
        "--cls_bs",
        type=int,
        default=1,
        help="Classification model inference batch size.")
    parser.add_argument(
        "--rec_bs",
        type=int,
        default=6,
        help="Recognition model inference batch size")
    parser.add_argument(
        "--backend",
        type=str,
        default="openvino",
        help="Type of inference backend, support ort/trt/paddle/openvino, default 'openvino' for cpu, 'tensorrt' for gpu"
    )

    return parser.parse_args()


def build_option(args):

    det_option = fd.RuntimeOption()
    cls_option = fd.RuntimeOption()
    rec_option = fd.RuntimeOption()

    if args.device.lower() == "gpu":
        det_option.use_gpu(args.device_id)
        cls_option.use_gpu(args.device_id)
        rec_option.use_gpu(args.device_id)

    if args.backend.lower() == "trt":
        assert args.device.lower(
        ) == "gpu", "TensorRT backend require inference on device GPU."
        det_option.use_trt_backend()
        cls_option.use_trt_backend()
        rec_option.use_trt_backend()

        # If use TRT backend, the dynamic shape will be set as follow.
        # We recommend that users set the length and height of the detection model to a multiple of 32.
        # We also recommend that users set the Trt input shape as follow.
        det_option.set_trt_input_shape("x", [1, 3, 64, 64], [1, 3, 640, 640],
                                       [1, 3, 960, 960])
        cls_option.set_trt_input_shape("x", [1, 3, 48, 10],
                                       [args.cls_bs, 3, 48, 320],
                                       [args.cls_bs, 3, 48, 1024])
        rec_option.set_trt_input_shape("x", [1, 3, 48, 10],
                                       [args.rec_bs, 3, 48, 320],
                                       [args.rec_bs, 3, 48, 2304])

        # Users could save TRT cache file to disk as follow.
        det_option.set_trt_cache_file(args.det_model + "/det_trt_cache.trt")
        cls_option.set_trt_cache_file(args.cls_model + "/cls_trt_cache.trt")
        rec_option.set_trt_cache_file(args.rec_model + "/rec_trt_cache.trt")

    elif args.backend.lower() == "pptrt":
        assert args.device.lower(
        ) == "gpu", "Paddle-TensorRT backend require inference on device GPU."
        det_option.use_paddle_infer_backend()
        det_option.paddle_infer_option.collect_trt_shape = True
        det_option.paddle_infer_option.enable_trt = True

        cls_option.use_paddle_infer_backend()
        cls_option.paddle_infer_option.collect_trt_shape = True
        cls_option.paddle_infer_option.enable_trt = True

        rec_option.use_paddle_infer_backend()
        rec_option.paddle_infer_option.collect_trt_shape = True
        rec_option.paddle_infer_option.enable_trt = True

        # If use TRT backend, the dynamic shape will be set as follow.
        # We recommend that users set the length and height of the detection model to a multiple of 32.
        # We also recommend that users set the Trt input shape as follow.
        det_option.set_trt_input_shape("x", [1, 3, 64, 64], [1, 3, 640, 640],
                                       [1, 3, 960, 960])
        cls_option.set_trt_input_shape("x", [1, 3, 48, 10],
                                       [args.cls_bs, 3, 48, 320],
                                       [args.cls_bs, 3, 48, 1024])
        rec_option.set_trt_input_shape("x", [1, 3, 48, 10],
                                       [args.rec_bs, 3, 48, 320],
                                       [args.rec_bs, 3, 48, 2304])

        # Users could save TRT cache file to disk as follow.
        det_option.set_trt_cache_file(args.det_model)
        cls_option.set_trt_cache_file(args.cls_model)
        rec_option.set_trt_cache_file(args.rec_model)

    elif args.backend.lower() == "ort":
        det_option.use_ort_backend()
        cls_option.use_ort_backend()
        rec_option.use_ort_backend()

    elif args.backend.lower() == "paddle":
        det_option.use_paddle_infer_backend()
        cls_option.use_paddle_infer_backend()
        rec_option.use_paddle_infer_backend()

    elif args.backend.lower() == "openvino":
        assert args.device.lower(
        ) == "cpu", "OpenVINO backend require inference on device CPU."
        det_option.use_openvino_backend()
        cls_option.use_openvino_backend()
        rec_option.use_openvino_backend()

    elif args.backend.lower() == "pplite":
        assert args.device.lower(
        ) == "cpu", "Paddle Lite backend require inference on device CPU."
        det_option.use_lite_backend()
        cls_option.use_lite_backend()
        rec_option.use_lite_backend()

    return det_option, cls_option, rec_option


args = parse_arguments()


det_model_file = os.path.join(args.det_model, "inference.pdmodel")
det_params_file = os.path.join(args.det_model, "inference.pdiparams")

cls_model_file = os.path.join(args.cls_model, "inference.pdmodel")
cls_params_file = os.path.join(args.cls_model, "inference.pdiparams")

rec_model_file = os.path.join(args.rec_model, "inference.pdmodel")
rec_params_file = os.path.join(args.rec_model, "inference.pdiparams")
rec_label_file = args.rec_label_file

det_option, cls_option, rec_option = build_option(args)

det_model = fd.vision.ocr.DBDetector(
    det_model_file, det_params_file, runtime_option=det_option)

cls_model = fd.vision.ocr.Classifier(
    cls_model_file, cls_params_file, runtime_option=cls_option)

rec_model = fd.vision.ocr.Recognizer(
    rec_model_file, rec_params_file, rec_label_file, runtime_option=rec_option)

# Parameters settings for pre and post processing of Det/Cls/Rec Models.
# All parameters are set to default values.
det_model.preprocessor.max_side_len = 960
det_model.postprocessor.det_db_thresh = 0.3
det_model.postprocessor.det_db_box_thresh = 0.6
det_model.postprocessor.det_db_unclip_ratio = 1.5
det_model.postprocessor.det_db_score_mode = "slow"
det_model.postprocessor.use_dilation = False
cls_model.postprocessor.cls_thresh = 0.9

# Create PP-OCRv3, if cls_model is not needed, just set cls_model=None .
ppocr_v3 = fd.vision.ocr.PPOCRv3(
    det_model=det_model, cls_model=cls_model, rec_model=rec_model)

# Set inference batch size for cls model and rec model, the value could be -1 and 1 to positive infinity.
# When inference batch size is set to -1, it means that the inference batch size
# of the cls and rec models will be the same as the number of boxes detected by the det model.
ppocr_v3.cls_batch_size = args.cls_bs
ppocr_v3.rec_batch_size = args.rec_bs

# Read the input image
im = cv2.imread(args.image)

# Predict and reutrn the results
result = ppocr_v3.predict(im)

print(result)

# Visuliaze the results.
vis_im = fd.vision.vis_ppocr(im, result)
cv2.imwrite("visualized_result.jpg", vis_im)
cv2.imshow("visualized_result.jpg", vis_im)
cv2.waitKey()
print("Visualized result save in ./visualized_result.jpg")

以上代码基于paddleocr的官方代码改进而来，故而可以参考官方的命令参数进行使用

```# 在CPU上使用Paddle Inference推理
python infer.py --det_model ch_PP-OCRv3_det_infer --cls_model ch_ppocr_mobile_v2.0_cls_infer --rec_model ch_PP-OCRv3_rec_infer --rec_label_file ppocr_keys_v1.txt --image 12.jpg --device cpu --backend paddle
# 在CPU上使用OenVINO推理
python infer.py --det_model ch_PP-OCRv3_det_infer --cls_model ch_ppocr_mobile_v2.0_cls_infer --rec_model ch_PP-OCRv3_rec_infer --rec_label_file ppocr_keys_v1.txt --image 12.jpg --device cpu --backend openvino
# 在CPU上使用ONNX Runtime推理
python infer.py --det_model ch_PP-OCRv3_det_infer --cls_model ch_ppocr_mobile_v2.0_cls_infer --rec_model ch_PP-OCRv3_rec_infer --rec_label_file ppocr_keys_v1.txt --image 12.jpg --device cpu --backend ort
# 在CPU上使用Paddle Lite推理
python infer.py --det_model ch_PP-OCRv3_det_infer --cls_model ch_ppocr_mobile_v2.0_cls_infer --rec_model ch_PP-OCRv3_rec_infer --rec_label_file ppocr_keys_v1.txt --image 12.jpg --device cpu --backend pplite
# 在GPU上使用Paddle Inference推理
python infer.py --det_model ch_PP-OCRv3_det_infer --cls_model ch_ppocr_mobile_v2.0_cls_infer --rec_model ch_PP-OCRv3_rec_infer --rec_label_file ppocr_keys_v1.txt --image 12.jpg --device gpu --backend paddle
# 在GPU上使用Paddle TensorRT推理
python infer.py --det_model ch_PP-OCRv3_det_infer --cls_model ch_ppocr_mobile_v2.0_cls_infer --rec_model ch_PP-OCRv3_rec_infer --rec_label_file ppocr_keys_v1.txt --image 12.jpg --device gpu --backend pptrt
# 在GPU上使用ONNX Runtime推理
python infer.py --det_model ch_PP-OCRv3_det_infer --cls_model ch_ppocr_mobile_v2.0_cls_infer --rec_model ch_PP-OCRv3_rec_infer --rec_label_file ppocr_keys_v1.txt --image 12.jpg --device gpu --backend ort
# 在GPU上使用Nvidia TensorRT推理
python infer.py --det_model ch_PP-OCRv3_det_infer --cls_model ch_ppocr_mobile_v2.0_cls_infer --rec_model ch_PP-OCRv3_rec_infer --rec_label_file ppocr_keys_v1.txt --image 12.jpg --device gpu --backend trt

代码运行效果图如下所示

程序运行时输出如下所示，可以看到大部分文字都准确识别了。

py 复制代码 
[INFO] fastdeploy/runtime/backends/openvino/ov_backend.cc(218)::fastdeploy::OpenVINOBackend::InitFromPaddle                                                            number of streams:1.
[INFO] fastdeploy/runtime/backends/openvino/ov_backend.cc(228)::fastdeploy::OpenVINOBackend::InitFromPaddle                                                            affinity:YES.
[INFO] fastdeploy/runtime/backends/openvino/ov_backend.cc(240)::fastdeploy::OpenVINOBackend::InitFromPaddle                                                            Compile OpenVINO model on device_name:CPU.
[INFO] fastdeploy/runtime/runtime.cc(286)::fastdeploy::Runtime::CreateOpenVINOBackend                                                                                  Runtime initialized with Backend::OPENVINO in Device::CPU.
[INFO] fastdeploy/runtime/backends/openvino/ov_backend.cc(218)::fastdeploy::OpenVINOBackend::InitFromPaddle                                                            number of streams:1.
[INFO] fastdeploy/runtime/backends/openvino/ov_backend.cc(228)::fastdeploy::OpenVINOBackend::InitFromPaddle                                                            affinity:YES.
[INFO] fastdeploy/runtime/backends/openvino/ov_backend.cc(240)::fastdeploy::OpenVINOBackend::InitFromPaddle                                                            Compile OpenVINO model on device_name:CPU.
[INFO] fastdeploy/runtime/runtime.cc(286)::fastdeploy::Runtime::CreateOpenVINOBackend                                                                                  Runtime initialized with Backend::OPENVINO in Device::CPU.
[INFO] fastdeploy/runtime/backends/openvino/ov_backend.cc(218)::fastdeploy::OpenVINOBackend::InitFromPaddle                                                            number of streams:1.
[INFO] fastdeploy/runtime/backends/openvino/ov_backend.cc(228)::fastdeploy::OpenVINOBackend::InitFromPaddle                                                            affinity:YES.
[INFO] fastdeploy/runtime/backends/openvino/ov_backend.cc(240)::fastdeploy::OpenVINOBackend::InitFromPaddle                                                            Compile OpenVINO model on device_name:CPU.
[INFO] fastdeploy/runtime/runtime.cc(286)::fastdeploy::Runtime::CreateOpenVINOBackend                                                                                  Runtime initialized with Backend::OPENVINO in Device::CPU.
det boxes: [[42,413],[483,391],[484,428],[43,450]]rec text: 上海斯格威铂尔大酒店 rec score:0.989301 cls label: 0 cls score: 1.000000
det boxes: [[187,456],[399,448],[400,480],[188,488]]rec text: 打浦路15号 rec score:0.994266 cls label: 0 cls score: 1.000000
det boxes: [[23,507],[513,488],[515,529],[24,548]]rec text: 绿洲仕格维花园公寓 rec score:0.984679 cls label: 0 cls score: 1.000000
det boxes: [[74,553],[427,542],[428,571],[75,582]]rec text: 打浦路252935号 rec score:0.996439 cls label: 0 cls score: 1.000000
Visualized result save in ./visualized_result.jpg
对文本切片进行识别
模型简介 模型名称 推荐场景 检测模型 方向分类器 识别模型
中英文超轻量PP-OCRv4模型（15.8M） ch_PP-OCRv4_xx 移动端&服务器端 推理模型 / 训练模型 推理模型 / 训练模型 推理模型 / 训练模型
中英文超轻量PP-OCRv3模型（16.2M） ch_PP-OCRv3_xx 移动端&服务器端 推理模型 / 训练模型 推理模型 / 训练模型 推理模型 / 训练模型
英文超轻量PP-OCRv3模型（13.4M） en_PP-OCRv3_xx 移动端&服务器端 推理模型 / 训练模型 推理模型 / 训练模型 推理模型 / 训练模型

各位可以自行截图生成如下图片，保存为13.jpg

识别代码如下所示，识别效果十分棒，各位可以换成自己的文本切片。

py 复制代码 
import fastdeploy as fd
import cv2
import os

rec_model_file = os.path.join('ch_ppocr_mobile_v2.0_cls_infer', "inference.pdmodel")
rec_params_file = os.path.join('ch_ppocr_mobile_v2.0_cls_infer', "inference.pdiparams")
rec_label_file = "ppocr_keys_v1.txt"

# Set the runtime option
rec_option =  fd.RuntimeOption()
rec_option.use_ort_backend()

# Create the rec_model
rec_model = fd.vision.ocr.Recognizer(
    rec_model_file, rec_params_file, rec_label_file, runtime_option=rec_option)

# Read the image
im = cv2.imread("13.jpg")

# Predict and return the result
result = rec_model.predict(im)
print(type(result),dir(result))
# User can infer a batch of images by following code.
# result = rec_model.batch_predict([im])

print(result)

识别结果如下所示，可以看到识别结果完全正确。

py 复制代码 
[INFO] fastdeploy/runtime/runtime.cc(300)::fastdeploy::Runtime::CreateOrtBackend        Runtime initialized with Backend::ORT in Device::CPU.
<class 'fastdeploy.libs.fastdeploy_main.vision.OCRResult'> ['__class__', '__delattr__', '__dir__', '__doc__', '__eq__', '__format__', '__ge__', '__getattribute__', '__gt__', '__hash__', '__init__', '__init_subclass__', '__le__', '__lt__', '__module__', '__ne__', '__new__', '__reduce__', '__reduce_ex__', '__repr__', '__setattr__', '__sizeof__', '__str__', '__subclasshook__', 'boxes', 'cls_labels', 'cls_scores', 'rec_scores', 'text']
rec text: 绿洲仕格维花园公寓 rec score:0.988077

3、其他功能

前文仅描述了常见功能的python使用，在faskdeploy还提供了c++部署案列，其python接口与c++接口基本一致。

除了图像分类、目标检测、语义分割、文本识别等功能外，faskdeploy还提供了人脸检测、人脸特征点、人脸识别、生成式模型、姿态检测、扣图、图像超分、目标跟踪等功能的部署案例。后续有使用到会在这里补充案例。

相关推荐
余生H17 分钟前
transformer.js（三）：底层架构及性能优化指南
javascript·深度学习·架构·transformer
罗小罗同学1 小时前
医工交叉入门书籍分享：Transformer模型在机器学习领域的应用｜个人观点·24-11-22
深度学习·机器学习·transformer
孤独且没人爱的纸鹤1 小时前
【深度学习】：从人工神经网络的基础原理到循环神经网络的先进技术，跨越智能算法的关键发展阶段及其未来趋势，探索技术进步与应用挑战
人工智能·python·深度学习·机器学习·ai
阿_旭1 小时前
TensorFlow构建CNN卷积神经网络模型的基本步骤：数据处理、模型构建、模型训练
人工智能·深度学习·cnn·tensorflow
羊小猪~~1 小时前
tensorflow案例7--数据增强与测试集, 训练集, 验证集的构建
人工智能·python·深度学习·机器学习·cnn·tensorflow·neo4j
极客代码1 小时前
【Python TensorFlow】进阶指南（续篇三）
开发语言·人工智能·python·深度学习·tensorflow
Seeklike1 小时前
11.22 深度学习-pytorch自动微分
人工智能·pytorch·深度学习
YRr YRr2 小时前
如何使用 PyTorch 实现图像分类数据集的加载和处理
pytorch·深度学习·分类
HPC_fac130520678165 小时前
以科学计算为切入点：剖析英伟达服务器过热难题
服务器·人工智能·深度学习·机器学习·计算机视觉·数据挖掘·gpu算力
老艾的AI世界13 小时前
AI翻唱神器，一键用你喜欢的歌手翻唱他人的曲目（附下载链接）
人工智能·深度学习·神经网络·机器学习·ai·ai翻唱·ai唱歌·ai歌曲
热门推荐
01(欧拉)openEuler系统添加网卡文件配置流程、(欧拉)openEuler系统手动配置ipv6地址流程、(欧拉)openEuler系统网络管理说明02PyTorch机器学习实现液态神经网络03【HarmonyOS】HUAWEI DevEco Studio 下载地址汇总04玄机平台应急响应—webshell查杀05Coze扣子平台完整体验和实践（附国内和国际版对比）06Ubuntu 20.04使用Livox mid 360 测试 FAST_LIO07【目标跟踪】相机运动补偿08Unity中PICO实现 隔空取物 和 接触抓取物体09RAG 实践- Ollama+RagFlow 部署本地知识库10怎样让音频速度变慢？请跟随以下方法进行操作