deep-saket/siamese_classification_oneshot

Object Classification using Siamese Net and Triplet Loss

1. This repo contains an implementation of siamese network using triplett loss for object identification.
   
2. The backbone used for this network in VGG19 having pretrained imagenet weights.
   
3. This model was build for research and experiment

Future Work

1. Accuracy, ROC curve to be implemented and integrated in the evaluation subroutine.
   
2. To be trained with more data and mode classes.
   
3. Add demo script

Dataset

1. Its a very small dataset with 83 classes
   
2. Link - https://drive.google.com/file/d/1UMuw0aletrs0X6Kdbr2hOEbzHoykANEY/view?usp=sharing

Saved models

Link - https://drive.google.com/file/d/1AYirOgPosRgOu8qehsDKj2FaziT3jVFc/view?usp=sharing

Dependencies

0. Python 3.8+
   
1. tensorflow > 2.10
   
2. cv2
   

Train

Change the parameters of run.sh and

source run.sh

Note - This repo is developed and maintained by Saket Mohanty