Wsinfer local model doc

docs/deep/wsinfer.md

@@ -1,6 +1,7 @@
-(wsinfer-extension)=
 # WSInfer
 
+(wsinfer-extension)=
+
 The [WSInfer QuPath extension](https://github.com/qupath/qupath-extension-wsinfer/) makes it possible to do patch-based deep learning inference for digital pathology, without any need for scripting.
 
 It's a collaboration between Stony Brook University ([WSInfer](https://wsinfer.readthedocs.io/en/latest/)) and the QuPath group (the [extension](https://github.com/qupath/qupath-extension-wsinfer/) that brings WSInfer models to QuPath).
@@ -26,7 +27,6 @@ A GPU is not required but can dramatically speed up processing.
 If you have an NVIDIA GPU and want to use it with WSInfer, you will need to install a version of CUDA compatible with PyTorch - please see {doc}`gpu`.
 :::
 
-
 ## Set-up
 
 With QuPath installed and running, drag and drop the WSInfer extension into the application and restart QuPath.
@@ -102,9 +102,9 @@ However the most (potentially) exciting additional option is the {guilabel}`Pref
 
 The options available will depend upon your computer's capabilities (at least as far as they could be discerned by Deep Java Library):
 
-* **CPU**: This is generally the safest - and slowest - option, because it should be supported on all computers.
-* **MPS**: This stands for *Metal Performance Shaders*, and should be available on recent Apple silicon - it is the Mac version of GPU acceleration
-* **GPU**: This should appear if you have an NVIDIA GPU, CUDA... and a little bit of luck.
+- **CPU**: This is generally the safest - and slowest - option, because it should be supported on all computers.
+- **MPS**: This stands for *Metal Performance Shaders*, and should be available on recent Apple silicon - it is the Mac version of GPU acceleration
+- **GPU**: This should appear if you have an NVIDIA GPU, CUDA... and a little bit of luck.
 
 If either MPS or GPU work for you, they should reduce the time required for inference by a *lot*.
 However configuration for GPU can be tricky, as it will depend upon other hardware and software on your computer - CUDA in particular.
@@ -119,6 +119,33 @@ It won't automatically find any existing PyTorch you might have installed: Deep
 If you have a compatible GPU, and want CUDA support, you'll need to ensure you have an appropriate CUDA installed *before* PyTorch is downloaded.
 :::
 
+## Using your own models
+
+In addition to models downloaded from the WSInfer zoo, you may also use your own models by putting them in a specific local directory called `user` or `local`. This in turn should be put into the model directory (the path can be seen in the WSI dialog) as a sibling of the `kaczmarj` directory that contains the downloaded models.
+
+Inside the local models directory, every model should be contained in a subdirectory. The subdirectory name becomes the model name in the model selection menu.
+Inside it, there will be:
+
+- the model in Torchscript format, with the name `torchscript_model.pt`, and
+- a `config.json` file, as explained [here](https://wsinfer.readthedocs.io/en/latest/user_guide.html#use-your-own-model). An example of config file is shown [here](https://github.com/SBU-BMI/wsinfer/issues/221).
+
+Thus, the structure of the models directory could be as follows:
+
+```text
+wsinfer
+ kaczmarj
+ breast-tumor-resnet34.tcga-brca
+ main
+ README.md
+ config.json
+ lfs-pointer.txt
+ torchscript_model.pt
+ local
+ my_model
+ config.json
+ torchscript_model.pt
+ wsinfer-zoo-registry.json
+```
 
 ## Scripting
 
@@ -137,11 +164,10 @@ where the `selectAnnotation()` line was added when I pressed the {guilabel}`Anno
 
 To process in batch, I would need to
 
-* Add my images to a QuPath project
-* Annotate the regions of interest in the images (and save the data)
-* Open the above script in QuPath's script editor
-* Choose {menuselection}`Run --> Run for project`, and select the images I want to process
-
+- Add my images to a QuPath project
+- Annotate the regions of interest in the images (and save the data)
+- Open the above script in QuPath's script editor
+- Choose {menuselection}`Run --> Run for project`, and select the images I want to process
 
 ## Identifying TILs (overlaying predictions of two models)