I have created two different options to process your duckweed image data. Note that they may provide different results from each other. Click on the option below that you would like to explore further
-
Download the standalone application and all files as a zip from the link here (note that this requires a utoronto account)
-
Extract all items into your computer and keep them in the same folder
-
Double click DuckPlate_Image_Processing_App_v2.exe to run the app. A window will appear. Do not close it while the app is running.
-
Once the app is ready and loaded, the following window will appear on your screen. Click on the button
- You can either choose one file or multiple. Note that the files must be sized 4056x3040, with up to 2 well plates
- Once you click Open, the main application will appear on the screen
- Press Run. You will be able to see the progress of the image processing. Once completed, you can view the results by clicking the button check results
This app will output the raw well images, binary masks, and data for each well into a folder named data in the same directory as your images
-
If you do not have MATLAB installed in your computer, go to MATLAB Runtime and install a runtime version compatible with your operating system, number R2022a (9.12). The installation may take a few minutes
-
Download the standalone application and all files as a zip from this link (note that this requires a utoronto account)
-
Extract all items into your computer and keep them in the same folder
-
To open the app, double click on the file Run_plate.exe.
-
A black window will appear. Keep it open whenever using the app, and wait for the main app to load
-
Skip set B and and go directly to step C below for more details
-
Open MATLAB in your computer
-
Prior to opening the image processing app on the command window, make sure that the current folder corresponds to the folder where the code is located at.
-
Next, write the following on the command window: Run_plate , or simply click open Run_plate.m file and press run
-
Go to step C below to view more details on how to run the app
tutorial-duckplate-app.mp4
- Once open, check the settings for the type of plate you will run. For more information, you can click on the headings for a description of each setting. Ensure that:
a. HSV color threshold settings are defined properly. The default ones may be from another run and could work well with your plate images
b. Select settings on the check boxes
c. Define plate dimensions
- You can then run one of the three protocols
Processing a single well plate
-
To process a single plate image, first select the desired well plate dimension (i.e. 24)
-
Click on RUN SINGLE PLATE, as shown below:
- You will be prompted to select an image. Make sure the selection is a raw image as shown below
- Next, a window will be displayed asking you to click anywhere on side you want to be selected. Here, I selected the plate on the left side (plate 1). Once you click, this window will automatically disappear.
- You can view the progress on the main app window. Wait until the app says completed. You can also click to view results on your computer
- You can continue processing other images or exit the app.
In case a batch of new images is taken, it may be necessary to update/re-train the machine learning models with new data for better accuracy. Although the model is written in python, retraining it should not require any previous Python knowledge. The training can be done in Google Colab. Follow the instructions for whichever model needs to be updated. Once done, download the model folder (.ckpt) and paste the unzipped version into the same directory as the python DuckPlate App.
96-well plate duckweed segmentation model
24-well plate duckweed segmentation model