<p>With the development of society and the economy, the demand for food nutrition evaluation is increasing. Consequently, various nutritional estimation methods have been proposed. However, there are still some issues that need to be further considered. Specifically, (1) Traditional methods often rely on specialized biochemical instruments and are typically confined to laboratory environments, thereby making them difficult to popularize; (2) Some vision-based methods only leverage RGB images as input, missing some necessary spatial information. To solve the above problems, we propose a novel Fusion and Bidirectional Feature Pyramid Network (FBFPN) for nutrition estimation. The proposed FBFPN is an end-to-end approach, which simultaneously takes RGB and depth images as input, the spatial information can be effectively utilized. Besides, we develop an RGB-D fusion module to excavate richer vision features. A multi-scale fusion module is proposed to fuse feature maps with different resolutions. Compared with state-of-the-art methods, the mean value of the PMAE for our method reaches 17.3<InlineEquation ID="IEq1"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="530_2025_1732_Article_IEq1.gif" Format="GIF" Height="16" Rendition="HTML" Resolution="72" Type="Linedraw" Width="15" /> </InlineMediaObject> <EquationSource Format="TEX">\(\%\)</EquationSource> <EquationSource Format="MATHML"><math> <mo>%</mo> </math></EquationSource> </InlineEquation>. The PMAE for calories, mass, carb and protein is 14.0<InlineEquation ID="IEq2"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="530_2025_1732_Article_IEq2.gif" Format="GIF" Height="16" Rendition="HTML" Resolution="72" Type="Linedraw" Width="15" /> </InlineMediaObject> <EquationSource Format="TEX">\(\%\)</EquationSource> <EquationSource Format="MATHML"><math> <mo>%</mo> </math></EquationSource> </InlineEquation>, 10.3<InlineEquation ID="IEq3"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="530_2025_1732_Article_IEq3.gif" Format="GIF" Height="16" Rendition="HTML" Resolution="72" Type="Linedraw" Width="15" /> </InlineMediaObject> <EquationSource Format="TEX">\(\%\)</EquationSource> <EquationSource Format="MATHML"><math> <mo>%</mo> </math></EquationSource> </InlineEquation>, 19.5<InlineEquation ID="IEq4"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="530_2025_1732_Article_IEq4.gif" Format="GIF" Height="16" Rendition="HTML" Resolution="72" Type="Linedraw" Width="15" /> </InlineMediaObject> <EquationSource Format="TEX">\(\%\)</EquationSource> <EquationSource Format="MATHML"><math> <mo>%</mo> </math></EquationSource> </InlineEquation> and 20.2<InlineEquation ID="IEq5"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="530_2025_1732_Article_IEq5.gif" Format="GIF" Height="16" Rendition="HTML" Resolution="72" Type="Linedraw" Width="15" /> </InlineMediaObject> <EquationSource Format="TEX">\(\%\)</EquationSource> <EquationSource Format="MATHML"><math> <mo>%</mo> </math></EquationSource> </InlineEquation>, respectively, representing improvements of 1.0<InlineEquation ID="IEq6"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="530_2025_1732_Article_IEq6.gif" Format="GIF" Height="16" Rendition="HTML" Resolution="72" Type="Linedraw" Width="15" /> </InlineMediaObject> <EquationSource Format="TEX">\(\%\)</EquationSource> <EquationSource Format="MATHML"><math> <mo>%</mo> </math></EquationSource> </InlineEquation>, 0.5<InlineEquation ID="IEq7"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="530_2025_1732_Article_IEq7.gif" Format="GIF" Height="16" Rendition="HTML" Resolution="72" Type="Linedraw" Width="15" /> </InlineMediaObject> <EquationSource Format="TEX">\(\%\)</EquationSource> <EquationSource Format="MATHML"><math> <mo>%</mo> </math></EquationSource> </InlineEquation>, 2.9<InlineEquation ID="IEq8"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="530_2025_1732_Article_IEq8.gif" Format="GIF" Height="16" Rendition="HTML" Resolution="72" Type="Linedraw" Width="15" /> </InlineMediaObject> <EquationSource Format="TEX">\(\%\)</EquationSource> <EquationSource Format="MATHML"><math> <mo>%</mo> </math></EquationSource> </InlineEquation> and 0.8<InlineEquation ID="IEq9"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="530_2025_1732_Article_IEq9.gif" Format="GIF" Height="16" Rendition="HTML" Resolution="72" Type="Linedraw" Width="15" /> </InlineMediaObject> <EquationSource Format="TEX">\(\%\)</EquationSource> <EquationSource Format="MATHML"><math> <mo>%</mo> </math></EquationSource> </InlineEquation>, respectively. These results demonstrate the effectiveness of our approach.</p>

错误:搜索内容不能为空,请输入英文关键词
错误:关键词超出字数限制,请精简
高级检索

Food nutrition estimation with RGB-D fusion module and bidirectional feature pyramid network

  • Boyuan Ma,
  • Donglin Zhang,
  • Xiao-Jun Wu

摘要

With the development of society and the economy, the demand for food nutrition evaluation is increasing. Consequently, various nutritional estimation methods have been proposed. However, there are still some issues that need to be further considered. Specifically, (1) Traditional methods often rely on specialized biochemical instruments and are typically confined to laboratory environments, thereby making them difficult to popularize; (2) Some vision-based methods only leverage RGB images as input, missing some necessary spatial information. To solve the above problems, we propose a novel Fusion and Bidirectional Feature Pyramid Network (FBFPN) for nutrition estimation. The proposed FBFPN is an end-to-end approach, which simultaneously takes RGB and depth images as input, the spatial information can be effectively utilized. Besides, we develop an RGB-D fusion module to excavate richer vision features. A multi-scale fusion module is proposed to fuse feature maps with different resolutions. Compared with state-of-the-art methods, the mean value of the PMAE for our method reaches 17.3 \(\%\) % . The PMAE for calories, mass, carb and protein is 14.0 \(\%\) % , 10.3 \(\%\) % , 19.5 \(\%\) % and 20.2 \(\%\) % , respectively, representing improvements of 1.0 \(\%\) % , 0.5 \(\%\) % , 2.9 \(\%\) % and 0.8 \(\%\) % , respectively. These results demonstrate the effectiveness of our approach.