For an efficient analysis of stomatal traits—like number, size, and closure rate—which are crucial for understanding plant responses to the environment, we introduce StomaVision. This automated tool simplifies stomatal detection and measurement on images from various plant species, including field samples. It enhances research efficiency by processing images quickly, even from videos, increasing the data available for robust analysis.
This tool is developed and maintained by the Bioinformatics and Computational Genomics Group at Academia Sinica, in collaboration with engineers from Instill AI.
To ensure maximum portability and ease of use, all necessary software packages and modules have been encapsulated within Docker containers.
Ting-Li Wu, Po-Yu Chen, Xiaofei Du, Heiru Wu, Jheng-Yang Ou, Po-Xing Zheng, Yu-Lin Wu, Ruei-Shiuan Wang, Te-Chang Hsu, Chen-Yu Lin, Wei-Yang Lin, Ping-Lin Chang, Chin-Min Kimmy Ho, Yao-Cheng Lin. StomaVision: stomatal trait analysis through deep learning
For the impatient users - online app
We have set up a public Streamlit app powered by Instill AI for you to try out, you can access it here
This public online version of StomaVision is hosted on Streamlit and powered by Instill Cloud. If the instance remains unused for an extended period, it may be shut down. Please note that relaunching the instance could take approximately 5 minutes.
NOTE: The recommended image size is a long axis of less than 2000 pixels (e.g., 1920x1080 or 1280x960 pixels). A magnification of 20X is recommended for the objective lens. Other magnifications may affect prediction accuracy. Multiple photos can be uploaded at once.
StomaVision was developed based on YOLOv7 in combination with the Instill Core and Streamlit. The whole environment is packaged in two main containers. To setup your local instance of StomaVision app, follow the steps below
Launch local instance of Instill Core
First you will need a local instance of Instill Core to host and serve the StomaVision model and utilize the versatile AI pipeline, refer to Local Instill Core guide to set it up
After you have Instill Core running locally, you can go to https://localhost:3000s pipeline builder page to play around with the new created pipeline, you can learn more about it here
Now we can launch the local instance of StomaVision streamlit app that connects to your local Instill Core instance, refer to Local Stremalit guide to set it up
You can also fine-tune and serve your own stomata detection model, please refer to TRAIN guide
Ting-Li Wu, Po-Yu Chen, Xiaofei Du, Heiru Wu, Jheng-Yang Ou, Po-Xing Zheng, Yu-Lin Wu, Ruei-Shiuan Wang, Te-Chang Hsu, Chen-Yu Lin, Wei-Yang Lin, Ping-Lin Chang, Chin-Min Kimmy Ho, Yao-Cheng Lin. StomaVision: stomatal trait analysis through deep learning