Orhcid Classification

The architecture of the model is as follows:

It include 2 part:

1. Segmentation orchid flower (SOF): detects and segments orchid flowers in image
2. Classifier orchid’s species (COS): classifies orchid flowers to its label

This is the result of our model:

Dataset	mAP	Weights
Orchid Flowers Dataset
Orchid Flowers Classification

Segmentation orchid flower (SOF)

It is trained by fine tune the yolact model

Yolact paper: https://arxiv.org/abs/1904.02689

The result of SOF is here:

Dataset	mAP	Weights
Orchid Flowers Dataset
Orchid Flowers Classification

Classifier orchid’s species (COS)

Based on idea of this paper

We extend it by increase the number of model it use to implement features extraction and change the algorithm it use to select features. This is the result of COS:

Dataset	mAP	Weights
Orchid Flowers Dataset
Orchid Flowers Classification