quarrying/quarrying-plant-id

植物识别 (Plant Recognition)

Features

• Identify 4,066 plant taxa (maybe genus, species, subspecies, variety, etc.).
• Model size is 29.5M; top1/top5 accuracy is 0.848/0.959.
• Open-source model, easy-to-use interface.
• Update continuously.

Installation

You need install Anaconda, then run below:

git clone https://github.com/quarrying/quarrying-plant-id.git
cd quarrying-plant-id
conda create -n plantid python=3.6 -y
conda activate plantid
pip install -r requirements.txt

Usage

Method I: Python Interface

import cv2
import plantid

plant_identifier = plantid.PlantIdentifier()
image = cv2.imread(image_filename)
outputs = plant_identifier.identify(image, topk=5)
if outputs['status'] == 0:
    print(outputs['results'])
else:
    print(outputs)

You can also see demo.py.

Method II: Website

Goto https://www.quarryman.cn/plant which powered by this repo.

Method III: Telegram

For those who use telegram, the algorithm is now deployed on Yuuka.

Details

See https://zhuanlan.zhihu.com/p/364346303.

ChangeLog

• 20211024 Release model which supports 4066 plant taxa, top1/top5 accuracy is 0.848/0.959.
• 20210905 Release model which supports 2759 plant taxa, top1/top5 accuracy is 0.890/0.971.
• 20210718 Release model which supports 2002 plant taxa.
• 20210609 Release model which supports 1630 plant taxa.
• 20210413 Release model which supports 1320 plant taxa.

quarrying-plant-id's Issues

本地运行和网页结果不一致

你好大佬，我下载了这个代码本地运行，然后同一张图的识别结果，跟你部署的网页服务的结果不一样，网页上的识别准确一点。是不是代码这里的结果集有点滞后了？

大佬有计划把环境升级到 py3.8 吗？弱弱地问下...

Could you share the dataset

Hi, I would like to train a model of plants, would you like to share you dataset, thanks