Please enable JavaScript.
Coggle requires JavaScript to display documents.
Food Photo Recogniton for Dietary Tracking: System and Experiment (Results…
Food Photo Recogniton for Dietary Tracking: System and Experiment
Aim
medical institution
track activity
image recognition
portion size
App name
Diet Lense
System architecture
environemnt
Doctor
interact with user
centralization
Medical Knowledge
Social media
User
Model
Dataset
categories
249
300 images\categorie
classification
RasNet
deep model
ResNet-50
50 convolutional layers
training
stochastic gradient desent
momentum
0.9
initial learning rate
0.01
Results
problems
chinese foods
Malay food
covered with sauces
dataset
top 1
68.1%
top 5
89.9%
popular foods
top 1
75.2%
top 5
93.1%
real user photos
top-1
53%
top 5
71%
logging experience
fastest among compared systems
11 sec