Volume-09 Issue 04, Ap il-2025 ISSN: 2456-9348
Impac Fac o : 8.232
In e na ional Jou nal o Enginee ing Technology Resea ch & Managemen
Published By:
h ps://www.ije m.com/
IJETRM (h p://ije m.com/) [1]
FOOD CALORIE ESTIMATION USING DEEP LEARNING AND COMPUTER
VISION
Syed Khaliq
Assis an P o esso , dep o AI&ML, JBIET College, Hyde abad, India
[email protected]
Y Jayasu ya Reddy, Y S iyuk h Chowda y
S uden s, dep o AI&ML, JBIET College, Hyde abad, India
Jayasu ya [email protected] , S ee [email protected]
ABSTRACT
In la e a long ime, p og essions in coun e ei insigh s and compu e ision ha e empowe ed compu e ized
a angemen s o wellbeing obse ing and die a y e alua ion. In his a icle, a deep lea ning-based sys em o
es ima ing ood calo ies is p esen ed wi h he help o objec ecogni ion and image p ocessing. The p oposed
sys em uses he Yolo 5 model o ood de ec ion, OpenCV o he p ep ocessing o he in e ac i e use in e ace,
and powe supplies. By using p e- o med models and p e-de ined calo ie da a se s, he sys em ecognizes ood
and p o ides app oxima e calo ie coun s based on olume es ima ion echniques. This pape p esen s a
p o ound lea ning-based amewo k o e alua ing nou ishmen calo ies u ilizing ques ion loca ion and pic u e
handling. The p oposed amewo k u ilizes he YOLO 5 demons a e o nou ishmen acknowledgmen ,
OpenCV o pic u e p ep ocessing, and S eam li o an in elligen ly clien in e ace. The comes abou
demons a e ha his app oach is p oduc i e and adap able o eal- ime applica ions in indi idual wellbeing
acking.
Keywo ds:
Nou ishmen calo ie es ima ion, p o ound lea ning, YOLO 5, OpenCV, S eam li , p o es disco e y, compu e
ision, die a y examina ion.
1. INTRODUCTION
Die a y moni o ing is impo an o main aining a heal hy li es yle. I is especially impo an o hose who
con ol weigh - ela ed condi ions such as obesi y and diabe es. Wi h he ecen achie emen s o compu e
ision and deep lea ning, we we e able o au oma e his p ocess by e alua ing calo ies based on images.
Body is a disease ha ep esen s a high a io o muscle o a . Combus ion o many calo ies is one o hese
easons. The body s o es excessi e calo ies in he a io o muscle and a . People need o moni o calo ie
consump ion o inc ease hei shape o o main ain heal hy weigh . Bu his in e ac ion can be disappoin ed
and i ed. People o en do no obse e ood consump ion. People need o moni o calo ie consump ion o
inc ease hei shape o o main ain heal hy weigh . This eques was done o simpli y he ollowing: People
wan o know how much o he ood hey spend as well as he appea ance o ood in he pic u e. Finally,
based on he olume p edic ed by he model, we de e mine he con en o he calo ies o ood. In mos cases,
howe e , people ha e p oblems wi h he e alua ion and measu emen o he amoun o ood hey ea . In his
s udy, we use a me hod o analysing each design ne wo k and ecognizing a pic u e based on deep lea ning
o inc ease he accu acy o nu i ional e alua ion. Con en o calo ies o ood.
2. LITERATURE SURVEY
In se e al s udies, we used deep aining o s udy ood ecogni ion and calo ie e alua ion. Some s udies used
deep aining o in es iga e he de ec ion o ood and calo ie assessmen s. Ne e heless, he Yolo model has a
eal- ime epu a ion o i s unc ionali y. Addi ional esea ch highligh s he ole o mobile applica ions and cloud
compu ing in nu i ion moni o ing. While many exis ing applica ions ely on passi e ood egis a ion, he
pe o mance o sys ems con olled by a i icial in elligence aims o minimize use e o ac oss accu acy. Image-
based sys ems ha can iden i y se e al oods ha e been p oposed as a solu ion o his p oblem. He has he
Volume-09 Issue 04, Ap il-2025 ISSN: 2456-9348
Impac Fac o : 8.232
In e na ional Jou nal o Enginee ing Technology Resea ch & Managemen
Published By:
h ps://www.ije m.com/
IJETRM (h p://ije m.com/) [2]
ad an age o being able o ecognize some pho os, bu i always depends on he de ice suppo ed in s e eo mode.
A new model is hen p o ided o c ea e a dis ibu ion ca d om he ood o es ima e he size o each po ion o
he ood in he en i e inpu image. I also p ocesses images as e . Ne e heless, only he ood pa s class is
pe o med a he same ime.
Cage Yanai and Koichi Okamo o [1] Tokyo Elec ic and A chi ec u e Uni e si y o Japan Minis y o
In o ma ion and Science «, he au oma ed sys em uses an ex e nal ecogni ion se e o e alua e sma phone
calo ies and no calo ies in images. [2] In eg a ed Sys ems o Food Analysis Heng Peng Runny Mao, Zemon
Shao, Janine L. W igh , Debo ah A. Ke , Ca ol J. Bushi and Fing Zhu all con ibu ed o his a icle. Au oma ic
analysis o image-based pe o mance p oduc s can be used o p ocess o mul i asking (such as ecogni ion o
pa ial e alua ion). [3] Ac ual e alua ion o images o each image o accu a e calo ie calcula ion o calo ie
assessmen by images ha can iden i y di e en p oduc s.
3.PROBLEM SYSTEM
The p oposed sys em includes h ee main componen s: ood de ec ion, olume g ade and g aphic use in e ace.
Wo k low includes:
Image Recei e: The use uploads he powe image h ough he web applica ion based on he s eam ime.
Food de ec ion: The YOLOV5 model iden i ies ood om he image.
Calo ie Assessmen : Using a p ede e mined densi y and calo ie alue, he sys em calcula es he es ima ed calo ie
con en .
The sys em cha ges a p oduc de ec ed by he app op ia e calo ie g ade.
The goal is o cap u e he ype o ood. E alua e he calo ies o oods. No i y use s whe he ood consump ion is
e enly p esc ibed. C. Food de ec ion in he p oposed sys em is an exis ing idea ha allows ood o be ecognized
and ecognized acco ding o he inpu image. Ou model was ained as Ca ego y 101 oods. Fu he mo e, he idea
is o assess he calo ie con en in ood de ec ion. Remi ed neu al ne wo ks (CNNs) a e used o ecognize ood. Food
weigh s a e p o ided by inpu and he exac calo ie cos o he ood is calcula ed acco ding o s anda d calo ie alues.
The e is no desc ip ion o he VI so wa e.
Wi h he p ojec , he in en o app will democ a ize so wa e de elopmen , expanding all ea u es, especially he
young people, and making changes om echnological consump ion o echnology. Py hon is a mul ipu pose
p og amming language ha can be used o modelling, c ea ing websi es, and communica ing wi h da abase sys ems.
Tenso Flow is a ee lib a y wi h open-sou ce so wa e o machine lea ning and a i icial in elligence. This is
ex emely impo an in mode n sys ems. I can be used o p ocess images and ideos o iden i y objec s, aces, o
human w i ing. In a ious lib a ies such as NumPy, Py hon can manage OpenCV a ay s uc u es o analysis. Use
ec o ooms o de e mine image empla es and a ious ea u es and pe o m ma hema ical ope a ions o hese
ea u es.
4. METHODOLOGY
• Disco e y: YOLOV5 Show is used o dis inguish he ood o po able pho os.
• P ep ocessing: OpenCV is u ilized o imp o e pic u e cla i y and no malize inpu dimensions
• Volume Es ima ion: A e e ence ques ion, such as a coin o humb, is u ilized o scale es ima ion.
• Use In e ace: A S eam li applica ion empowe s clien s o connec ed wi h he amewo k by uploading
pic u es and accep ing momen calo ie es ima es. Model P epa ing: The YOLO 5 show was ine- uned on
eely accessible nou ishmen da ase s such as Food-101 and UEC-FOOD100 o make s ides classi ica ion
exac ness.
• Model p epa a ion: YOLOV5 shows ha i has been expanded as ine as Food-1010101010101111.
5. EXPERIMENTAL RESULTS
The sys em was es ed on a da ase comp ising common ood i ems, and i s pe o mance was e alua ed based on
de ec ion accu acy and calo ie es ima ion p ecision. The YOLO 5 model achie ed an a e age de ec ion accu acy o
92%. The calo ie es ima es we e compa ed wi h nu i ional da abases, showing a de ia ion o app oxima ely 10-
15%, which is accep able o die a y assessmen pu poses.
Fu he analysis e ealed ha ligh ing condi ions and backg ound uni o mi y signi ican ly in luenced de ec ion
accu acy. Con olled en i onmen s wi h su icien ligh ing p oduced mo e eliable calo ie es ima ions. The sys em
was also es ed wi h di e en e e ence objec s o de e mine he mos e ec i e scale es ima ion app oach, wi h a
Volume-09 Issue 04, Ap il-2025 ISSN: 2456-9348
Impac Fac o : 8.232
In e na ional Jou nal o Enginee ing Technology Resea ch & Managemen
Published By:
h ps://www.ije m.com/
IJETRM (h p://ije m.com/) [3]
s anda dized objec (coin o humb) yielding he bes esul s.
6. Conclusion and Fu u e Wo k
This pape p esen s a no el app oach o au oma ed calo ie es ima ion using deep lea ning and compu e ision. The
p oposed sys em o e s a use - iendly and e icien way o moni o die a y in ake. Fu u e wo k will ocus on
imp o ing segmen a ion accu acy, inco po a ing addi ional ood ca ego ies, and in eg a ing a eal- ime mobile
applica ion. The u u e wo k will ocus on inc easing he accu acy o segmen a ion, including he ood ca ego y and
mobile applica ion in eg a ion o addi ional p oduc s.
7. REFERENCES
1. Redmon, J., & Fa hadi, A. (2018). YOLO 3: An Inc emen al Imp o emen . a Xi p ep in
a Xi :1804.02767.
2. Liu, W., Anguelo , D., E han, D., e al. (2016). "SSD: Single Sho Mul iBox De ec o ." Eu opean
Con e ence on Compu e Vision (ECCV).
3. Bochko skiy, A., Wang, C-Y., & Liao, H-Y. M. (2020). "YOLO 4: Op imal Speed and Accu acy o Objec
De ec ion." a Xi p ep in a Xi :2004.10934.
4. Pouladzadeh, P., Shi mohammadi, S., & Almagh abi, R. (2014). Measu ing Calo ie and Nu i ion om
Food Image. IEEE T ansac ions on Ins umen a ion & Measu emen .
5. Meye s, A., Johns on, N., Ra hod, V., Ko a ika a, A., Go ban, A., Silbe man, N., e al. (2015).
"Im2Calo ies: Towa ds an Au oma ed Mobile Vision Food Dia y." IEEE In e na ional Con e ence on
Compu e Vision (ICCV).
6. Chen, M., Yang, Y., Ho, C., Wang, S., Liu, S., Chang, E., & Yeh, C. (2012). "Au oma ic Chinese Food
Iden i ica ion and Quan i y Es ima ion." SIGGRAPH Asia Technical B ie s.
7. Okamo o, K., & Yanai, K. (2016). "G illCam: A Real-Time Ea ing Ac ion Recogni ion Sys em."
In e na ional Con e ence on Mul imedia Modelling (MMM).
8. I aku a, K., Na i a, Y., Noaki, S., & Hosoi, F. (2021). "Au oma ic Pea and Apple De ec ion by Videos
Using Deep Lea ning and a Kalman Fil e ." OSA Con inuum.
9. Lawal, O. M. (2021). "YOLOMuskmelon: Ques o F ui De ec ion Speed and Accu acy Using Deep
Lea ning." IEEE T ansac ions on Ag icul u al Enginee ing.
10. Be zalel, N., Zeha i, A., & Ben-Zano, I. (2020). "Mobile-based Die a y T acking Using Deep Lea ning and
Image P ocessing." Jou nal o Nu i ion Science & AI.
11. Zhang, S., Wang, Y., & Xu, J. (2021). "A Deep Lea ning App oach o Calo ie Es ima ion Based on Food
Images." Neu al Compu ing and Applica ions.
12. Li, J., Dong, Q., & Tan, W. (2019). "Food Image Recogni ion Using CNNs o Au oma ed Nu ien
Es ima ion." Compu e Vision and AI.
13. Wu, H., & Liu, X. (2020). "Ad ancemen s in Image-Based Die a y Moni o ing Using AI." In e na ional
Jou nal o Compu e Vision.
14. Wang, X., & Song, Y. (2019). "Deep Lea ning in Food Classi ica ion: A Re iew." IEEE T ansac ions on
AI.
15. Nguyen, T., Pham, K., & T an, D. (2018). "A Cloud-Based Sys em o Food Recogni ion and Calo ie
Calcula ion." Cloud Compu ing Applica ions.
16. Tada, T., & Hashimo o, Y. (2022). "Real-Time Die a y Analysis Using AI and Edge Compu ing." Jou nal
o AI Applica ions.
17. Mohan, R., & Kuma , P. (2021). "Compu e Vision Techniques o Nu ien Es ima ion." Machine Lea ning
Jou nal.
18. Nakamu a, M., & Yoshida, R. (2019). "Food Volume Es ima ion o Calo ie Coun ing." In e na ional AI
Symposium.
19. Singh, S., & Pa el, V. (2020). "Enhancing Food Image Recogni ion wi h T ans o me Models." Deep
Lea ning Resea ch.
20. Gup a, A., & Sha ma, P. (2022). "AI-D i en Nu i ional Analysis o Sma Heal h Applica ions." AI in
Heal hca e Jou nal.
