图片格式的问题 - 地平线开发者社区

自己的模型 lenet---三通道

在运行03_build.sh时候出现错误，以下是log日志

----------------------------------------------------------------------------------------------------------------------------------------------------------

2020-12-10 19:59:21,772 INFO hb_mapper_makertbin 54 Working dir: /data/x3/x3_tc_1.1.17e/x3-toolchain-1.1.17e.tar/horizon_x3_tc_1.1.17e/samples/03_classification/00_lenet/mapper/model_output

2020-12-10 19:59:21,772 INFO hb_mapper_makertbin 55 Start Model Convert....

2020-12-10 19:59:21,773 DEBUG helper 118 Model input names: ['input.1']

2020-12-10 19:59:21,776 DEBUG mapper_conf_parser 468 mean value: [0. 0. 0.]

2020-12-10 19:59:21,952 DEBUG mapper_conf_parser 532 layer run on cpu:

[]

2020-12-10 19:59:21,952 DEBUG hb_mapper_makertbin 280 call build params:

{'march': 'bernoulli2', 'debug_mode': False, 'save_model': True, 'name_prefix': 'lenet', 'input_dict': {'input.1': {'input_shape': [1, 3, 32, 32], 'expected_input_type': 'RGB_128', 'original_input_type': 'BGR', 'means': array([0., 0., 0.], dtype=float32)}}, 'cali_dict': {'calibration_type': 'kl', 'calibration_loader': {'input.1': <horizon_nn.data.loader.TransformLoader object at 0x7f0af20b2a58>}, 'per_channel': False, 'max_percentile': 1.0}, 'hbdk_dict': {'compile_mode': 'latency', 'debug': False, 'optimize_level': 'O3', 'input_source': {'input.1': 'ddr'}}, 'layer_dict': {'run_on_cpu': []}}

2020-12-10 19:59:21,963 INFO build 31 [Thu Dec 10 19:59:21 2020] Start to Horizon NN Model Convert.

2020-12-10 19:59:21,963 INFO build 133 Parsing the input parameter:{'input.1': {'input_shape': [1, 3, 32, 32], 'expected_input_type': 'RGB_128', 'original_input_type': 'BGR', 'means': array([0., 0., 0.], dtype=float32)}}

2020-12-10 19:59:21,964 INFO build 146 Parsing the calibration parameter

2020-12-10 19:59:21,969 DEBUG Image 405 Importing BlpImagePlugin

2020-12-10 19:59:21,972 DEBUG Image 405 Importing BmpImagePlugin

2020-12-10 19:59:21,972 DEBUG Image 405 Importing BufrStubImagePlugin

2020-12-10 19:59:21,973 DEBUG Image 405 Importing CurImagePlugin

2020-12-10 19:59:21,973 DEBUG Image 405 Importing DcxImagePlugin

2020-12-10 19:59:21,974 DEBUG Image 405 Importing DdsImagePlugin

2020-12-10 19:59:21,974 DEBUG Image 405 Importing EpsImagePlugin

2020-12-10 19:59:21,975 DEBUG Image 405 Importing FitsStubImagePlugin

2020-12-10 19:59:21,976 DEBUG Image 405 Importing FliImagePlugin

2020-12-10 19:59:21,976 DEBUG Image 405 Importing FpxImagePlugin

2020-12-10 19:59:22,027 DEBUG Image 405 Importing FtexImagePlugin

2020-12-10 19:59:22,027 DEBUG Image 405 Importing GbrImagePlugin

2020-12-10 19:59:22,028 DEBUG Image 405 Importing GifImagePlugin

2020-12-10 19:59:22,028 DEBUG Image 405 Importing GribStubImagePlugin

2020-12-10 19:59:22,028 DEBUG Image 405 Importing Hdf5StubImagePlugin

2020-12-10 19:59:22,028 DEBUG Image 405 Importing IcnsImagePlugin

2020-12-10 19:59:22,030 DEBUG Image 405 Importing IcoImagePlugin

2020-12-10 19:59:22,030 DEBUG Image 405 Importing ImImagePlugin

2020-12-10 19:59:22,031 DEBUG Image 405 Importing ImtImagePlugin

2020-12-10 19:59:22,032 DEBUG Image 405 Importing IptcImagePlugin

2020-12-10 19:59:22,032 DEBUG Image 405 Importing JpegImagePlugin

2020-12-10 19:59:22,033 DEBUG Image 405 Importing Jpeg2KImagePlugin

2020-12-10 19:59:22,033 DEBUG Image 405 Importing McIdasImagePlugin

2020-12-10 19:59:22,033 DEBUG Image 405 Importing MicImagePlugin

2020-12-10 19:59:22,033 DEBUG Image 405 Importing MpegImagePlugin

2020-12-10 19:59:22,034 DEBUG Image 405 Importing MpoImagePlugin

2020-12-10 19:59:22,034 DEBUG Image 405 Importing MspImagePlugin

2020-12-10 19:59:22,035 DEBUG Image 405 Importing PalmImagePlugin

2020-12-10 19:59:22,036 DEBUG Image 405 Importing PcdImagePlugin

2020-12-10 19:59:22,036 DEBUG Image 405 Importing PcxImagePlugin

2020-12-10 19:59:22,036 DEBUG Image 405 Importing PdfImagePlugin

2020-12-10 19:59:22,049 DEBUG Image 405 Importing PixarImagePlugin

2020-12-10 19:59:22,049 DEBUG Image 405 Importing PngImagePlugin

2020-12-10 19:59:22,049 DEBUG Image 405 Importing PpmImagePlugin

2020-12-10 19:59:22,050 DEBUG Image 405 Importing PsdImagePlugin

2020-12-10 19:59:22,050 DEBUG Image 405 Importing SgiImagePlugin

2020-12-10 19:59:22,050 DEBUG Image 405 Importing SpiderImagePlugin

2020-12-10 19:59:22,051 DEBUG Image 405 Importing SunImagePlugin

2020-12-10 19:59:22,051 DEBUG Image 405 Importing TgaImagePlugin

2020-12-10 19:59:22,052 DEBUG Image 405 Importing TiffImagePlugin

2020-12-10 19:59:22,052 DEBUG Image 405 Importing WebPImagePlugin

2020-12-10 19:59:22,053 DEBUG Image 405 Importing WmfImagePlugin

2020-12-10 19:59:22,054 DEBUG Image 405 Importing XbmImagePlugin

2020-12-10 19:59:22,055 DEBUG Image 405 Importing XpmImagePlugin

2020-12-10 19:59:22,056 DEBUG Image 405 Importing XVThumbImagePlugin

2020-12-10 19:59:22,229 DEBUG tool_utils 94 exception in makertbin

2020-12-10 19:59:22,231 DEBUG tool_utils 95 Traceback (most recent call last):

File "/root/.local/lib/python3.6/site-packages/x3_tc_ui/utils/tool_utils.py", line 92, in __decorator

func(*args, **kargs)

File "/root/.local/lib/python3.6/site-packages/x3_tc_ui/hb_mapper.py", line 86, in makertbin

MakertbinRunner(config, model_type).run(version)

File "/root/.local/lib/python3.6/site-packages/x3_tc_ui/hb_mapper_makertbin.py", line 322, in run

onnx_file=self.mp_conf.get_onnx_model(), **build_params)

File "/root/.local/lib/python3.6/site-packages/horizon_nn/build.py", line 778, in build_onnx

debug_mode=debug_mode)

File "/root/.local/lib/python3.6/site-packages/horizon_nn/build.py", line 99, in __init__

self.parse_cali_dict(cali_dict)

File "/root/.local/lib/python3.6/site-packages/horizon_nn/build.py", line 159, in parse_cali_dict

self.calibration_data = self.get_all_data(calibration_loader)

File "/root/.local/lib/python3.6/site-packages/horizon_nn/build.py", line 349, in get_all_data

datasets = gen_datasets(new_generators)

File "/root/.local/lib/python3.6/site-packages/horizon_nn/build.py", line 335, in gen_datasets

for all_data in zip(*generator):

File "/root/.local/lib/python3.6/site-packages/horizon_nn/data/loader.py", line 32, in __next__

return self.perform()

File "/root/.local/lib/python3.6/site-packages/horizon_nn/data/loader.py", line 35, in perform

data = self._parant.perform()

File "/root/.local/lib/python3.6/site-packages/x3_tc_ui/data/loader.py", line 42, in perform

return self.image_read_method(next(self.file_gen))

File "/root/.local/lib/python3.6/site-packages/x3_tc_ui/data/loader.py", line 51, in image_read_method

image = skimage.img_as_float(skimage.io.imread(file)).astype(

File "/root/.local/lib/python3.6/site-packages/skimage/io/_io.py", line 48, in imread

img = call_plugin('imread', fname, plugin=plugin, **plugin_args)

File "/root/.local/lib/python3.6/site-packages/skimage/io/manage_plugins.py", line 210, in call_plugin

return func(*args, **kwargs)

File "/root/.local/lib/python3.6/site-packages/skimage/io/_plugins/imageio_plugin.py", line 10, in imread

return np.asarray(imageio_imread(*args, **kwargs))

File "/root/.local/lib/python3.6/site-packages/imageio/core/functions.py", line 265, in imread

reader = read(uri, format, "i", **kwargs)

File "/root/.local/lib/python3.6/site-packages/imageio/core/functions.py", line 182, in get_reader

"Could not find a format to read the specified file in %s mode" % modename

ValueError: Could not find a format to read the specified file in single-image mode

2020-12-10 19:59:22,231 ERROR tool_utils 96 Could not find a format to read the specified file in single-image mode

2020-12-10 19:59:22,233 INFO build 34 [Thu Dec 10 19:59:22 2020] End to Horizon NN Model Convert.

-------------------------------------------------------------------------------------------------------------------------------------------------------------------

我的模型的配置文件

# The material in this file is confidential and contains trade secrets

# of Horizon Robotics Inc. This is proprietary information owned by

# Horizon Robotics Inc. No part of this work may be disclosed,

# reproduced, copied, transmitted, or used in any way for any purpose,

# without the express written permission of Horizon Robotics Inc.

# 模型转化相关的参数

model_parameters:

# Caffe浮点网络数据模型文件

onnx_model: '../../../01_common/modelzoo/mapper/classification/lenet_onnx/lenet.onnx'

# Caffe网络描述文件

# prototxt: '../../../01_common/modelzoo/mapper/classification/resnet18/resnet18_deploy.prototxt'

# 指定模型转换过程中是否输出各层的中间结果，如果为True，则输出所有层的中间输出结果，

layer_out_dump: False

# 用于设置上板模型输出的layout, 支持NHWC和NCHW, 输入None则使用模型默认格式

output_layout: None

# 日志文件的输出控制参数，

# debug输出模型转换的详细信息

# info只输出关键信息

# warn输出警告和错误级别以上的信息

log_level: 'debug'

# 模型转换输出的结果的存放目录

working_dir: 'model_output'

# 模型转换输出的用于上板执行的模型文件的名称前缀

output_model_file_prefix: 'lenet'

# 模型输入相关参数, 若输入多个节点, 则应使用';'进行分隔, 使用默认缺省设置则写None

input_parameters:

# (可不填) 模型输入的节点名称, 此名称应与模型文件中的名称一致, 否则会报错, 不填则会使用模型文件中的节点名称

input_name: 'input.1'

# 网络实际执行时，输入给网络的数据格式，包括 nv12/rgbp/bgrp/yuv444_128/gray/featuremap,

# 如果输入的数据为yuv444_128, 模型训练用的是bgrp，则hb_mapper将自动插入YUV到BGRP(NCHW)转化操作

input_type_rt: 'rgbp'

# 网络训练时输入的数据格式，可选的值为rgbp/bgrp/gray/featuremap/yuv444_128

input_type_train: 'bgrp'

# 网络输入的预处理方法，主要有以下几种：

# no_preprocess 不做任何操作

# mean_file 减去从通道均值文件(mean_file)得到的均值

# data_scale 对图像像素乘以data_scale系数

# mean_file_and_scale 减去通道均值后再乘以scale系数

norm_type: 'mean_file'

# (可不填) 模型网络的输入大小, 以'x'分隔, 不填则会使用模型文件中的网络输入大小

input_shape: '1x3x32x32'

# 图像减去的均值存放文件, 文件内存放的如果是通道均值，均值之间必须用空格分隔

mean_file: './meanfile.txt'

# 图像预处理缩放比例，该数值应为浮点数

scale_value: 1.0

calibration_parameters:

# 模型量化的参考图像的存放目录，图片格式支持Jpeg、Bmp等格式，输入的图片

# 应该是使用的典型场景，一般是从测试集中选择20~50张图片，另外输入

# 的图片要覆盖典型场景，不要是偏僻场景，如过曝光、饱和、模糊、纯黑、纯白等图片

# 若有多个输入节点, 则应使用';'进行分隔

cal_data_dir: './calibration_data_bgrp'

# 如果输入的图片文件尺寸和模型训练的尺寸不一致时，并且preprocess_on为true，

# 则将采用默认预处理方法(opencv resize)，

# 将输入图片缩放或者裁减到指定尺寸，否则，需要用户提前把图片处理为训练时的尺寸

preprocess_on: True

# 模型量化的算法类型，支持kl、max，通常采用KL即可满足要求

calibration_type: 'kl'

# 模型的量化校准方法设置为promoter，mapper会根据calibration的数据对模型进行微调从而提高精度，

# promoter_level的级别，可选的参数为-1到2，建议按照0到2的顺序实验，满足精度即可停止实验

# -1: 不进行promoter

# 0：表示对模型进行轻微调节，精度提高比较小

# 1：表示相对0对模型调节幅度稍大，精度提高也比较多

# 2：表示调节比较激进，可能造成精度的大幅提高也可能造成精度下降

promoter_level: -1

# 编译器相关参数

compiler_parameters:

# 编译策略，支持bandwidth和latency两种优化模式;

# bandwidth以优化ddr的访问带宽为目标；

# latency以优化推理时间为目标

compile_mode: 'latency'

# 设置debug为True将打开编译器的debug模式，能够输出性能仿真的相关信息，如帧率、DDR带宽占用等

debug: False

# 编译模型指定核数，不指定默认编译单核模型, 若编译双核模型，将下边注释打开即可

# core_num: 2

# 优化等级可选范围为O0~O3

# O0不做任何优化, 编译速度最快，优化程度最低,

# O1-O3随着优化等级提高，预期编译后的模型的执行速度会更快，但是所需编译时间也会变长。

# 推荐用O2做最快验证

optimize_level: 'O3'