DeepImageJ

DeepImageJ brings pretrained neural networks into ImageJ and Fiji for life-science researchers who want to analyze microscopy images without writing code or learning how to train models. It runs image-to-image predictions on your own computer, using a standard CPU or a compatible GPU.

The available models cover cell segmentation, fluorescence super-resolution, density estimation and virtual staining. Examples include U-Net models for glioblastoma and HeLa segmentation, FRU-Net for small extracellular vesicles, and models that predict fluorescent labels from light-transmission images. The result you can produce depends on the model you choose.

Its connection to the BioImage Model Zoo gives researchers access to shared trained models in a compatible format. DeepImageJ also supports analysis pipelines with multiple deep learning steps, even when those steps use different frameworks. That matters if your microscopy workflow needs several models within the same ImageJ or Fiji environment.

The plugin supports PyTorch, TensorFlow and ONNX models. It runs on Linux, Windows and macOS, though framework compatibility differs between Intel and Apple Silicon Macs. GPU prediction is available for supported PyTorch and TensorFlow engines on Linux and Windows. ImgLib2 support lets it handle larger images.

The plugin is open source under the BSD 2-Clause license. Downloaded models have their own licences, and first-time model and engine setup requires downloads.