NU-InNet

This implements training of NU-InNet (Naresuan University Inception Network) from NU-InNet: Thai Food Image Recognition Using Convolutional Neural Networks on Smartphone by Chakkrit Termritthikun, et. al.

Module architecture of Inception (left), NU-Inception-1.0 (middle), and NU-Inception-1.1 (right)

GoogLeNet, NU-InNet 1.0, and NU-InNet 1.1 architectures

TH-FOOD50

Download TH-FOOD50 : https://github.com/chakkritte/THFOOD-50

Examples of Thai food images in the THFOOD-50 dataset

Result NU-InNet with TH-FOOD50

Single-crop (224x224) average of 10-fold-cross-validation

Network	Top-1 Accuracy	Top-5 Accuracy	Average Forward-Backward (ms/images)	Parameters (x10^6)
AlexNet	58.1	86.4	13.90	58.48
SqueezeNet	58.2	87.4	24.53	0.75
GoogLeNet	68.4	91.7	40.13	10.45
NU-InNet 1.0	69.8	92.3	18.16	0.88
NU-InNet 1.1	68.7	92.3	36.52	0.89

License and Citation

NU-InNet and TH-FOOD50 for non-commercial research/educational use.

Please cite NU-InNet and TH-FOOD50 in your publications if it helps your research:

( Thai ) NU-InNet และ TH-FOOD50 อนุญาตให้ใช้เฉพาะเพื่อการศึกษาและวิจัยเท่านั้น ห้ามนำไปใช้เชิงการค้าทุกรูปแบบ

หากคุณนำ NU-InNet หรือ TH-FOOD50 ไปใช้ในงานวิจัย กรุณาอ้างอิง

@article{termritthikun2017nu,
  title={NU-InNet: Thai Food Image Recognition Using Convolutional Neural Networks on Smartphone},
  author={Termritthikun, Chakkrit and Muneesawang, Paisarn and Kanprachar, Surachet},
  journal={Journal of Telecommunication, Electronic and Computer Engineering (JTEC)},
  volume={9},
  number={2-6},
  pages={63--67},
  year={2017}
}