Focus Quality Assessment for Fluorescence Images

Model Description

FocusCLIP is a CLIP-based model designed for evaluating the quality of images captured by an fluorescence microscope.
The model takes an image together with stain information (as text) and predicts how well the image meets quality standards.
The model was trained on microscope images captured at a 10× magnification.

How to Use

Step 1. Select Stain Information

• Choose the appropriate stain type from the dropdown menu.
• Only stain information available in the provided Dropbox list is supported.
• Custom or user-defined stain entries are not allowed.
• If your desired stain option is not listed, please select the stain with the closest wavelength.
• Maximal Excitation Wavelength (nm):
  • DAPI or Hoechst: 405
  • Alexa 488, FITC: 488
  • Alexa 546, Cy3: 546
  • Alexa 647, Cy5: 647

Step 2. Upload an Image

• Upload the microscope image you want the model to evaluate.
• Please note: uploaded images may be stored and used for research purposes to improve and validate the model.

Step 3. Run Inference

• Once everything is ready, click Submit to start the inference process.
• Processing may take a few seconds depending on system load.

Step 4. Review and Download Results

• The model output will be displayed directly below the interface.
• The focus quality will be categorized into Good, Moderate, or Bad.
• If needed, you can download the results as a CSV file for further analysis or record keeping.
• Example images for each category are additionally provided below to help you understand the evaluation criteria.

Example Images

The following examples illustrate focus quality (Good, Moderate, Bad) for four different stains.