comparison pytorch_embedding.xml @ 1:84f96c952c2c draft default tip

planemo upload for repository https://github.com/goeckslab/gleam.git commit 5b6cd961948137853177b14b0fff80a5d40e8a07

0:38333676a029

            <option value="efficientnet_b6">EfficientNet-B6</option>
            <option value="efficientnet_b7">EfficientNet-B7</option>
            <option value="efficientnet_v2_s">EfficientNetV2-S</option>
            <option value="efficientnet_v2_m">EfficientNetV2-M</option>
            <option value="efficientnet_v2_l">EfficientNetV2-L</option>
            <option value="googlenet">GoogLeNet</option>
            <option value="inception_v3">Inception-V3</option>
            <option value="mnasnet0_5">MNASNet-0.5</option>
            <option value="mnasnet0_75">MNASNet-0.75</option>
            <option value="mnasnet1_0">MNASNet-1.0</option>

                    <has_text text="sample_name" />
                    <has_n_columns min="1" />
                </assert_contents>
            </output>
        </test>
    </tests>
    <help>
        <![CDATA[
        **What it does**
        This tool extracts image embeddings using a selected deep learning model.

        **Inputs**
        - A zip file containing images to process.
        - A model selection for embedding extraction.
        - An option to apply normalization to the extracted embeddings.
        - A choice of image transformation type before processing.

        **Outputs**
        - A CSV file containing embeddings. Each row corresponds to an image, with the file name in the first column and embedding vectors in the subsequent columns.
        ]]>
    </help>

1:84f96c952c2c

            <option value="efficientnet_b6">EfficientNet-B6</option>
            <option value="efficientnet_b7">EfficientNet-B7</option>
            <option value="efficientnet_v2_s">EfficientNetV2-S</option>
            <option value="efficientnet_v2_m">EfficientNetV2-M</option>
            <option value="efficientnet_v2_l">EfficientNetV2-L</option>
            <option value="gpfm">GPFM (Generalizable Pathology Foundation Model)</option>
            <option value="googlenet">GoogLeNet</option>
            <option value="inception_v3">Inception-V3</option>
            <option value="mnasnet0_5">MNASNet-0.5</option>
            <option value="mnasnet0_75">MNASNet-0.75</option>
            <option value="mnasnet1_0">MNASNet-1.0</option>

                    <has_text text="sample_name" />
                    <has_n_columns min="1" />
                </assert_contents>
            </output>
        </test>
        <test>
            <param name="input_zip" value="1_digit.zip" ftype="zip" />
            <param name="model_name" value="gpfm" />
            <param name="apply_normalization" value="true" />
            <param name="transform_type" value="RGB" />
            <output name="output_csv">
                <assert_contents>
                    <has_text text="sample_name" />
                    <has_n_columns min="1" />
                </assert_contents>
            </output>
        </test>
    </tests>
    <help>
        <![CDATA[
        **What it does**
        This tool extracts image embeddings using a selected deep learning model, including specialized pathology models like GPFM.

        **Inputs**
        - A zip file containing images to process.
        - A model selection for embedding extraction (includes GPFM for pathology images).
        - An option to apply normalization to the extracted embeddings.
        - A choice of image transformation type before processing.

        **Models Available**
        - Standard computer vision models (ResNet, EfficientNet, ViT, etc.)
        - GPFM: Generalizable Pathology Foundation Model - specialized for medical/pathology images
          * Automatically downloads 1.2GB pretrained weights on first use
          * Uses DinoVisionTransformer architecture (1024-dimensional embeddings)
          * Optimized for histopathology images at 224x224 resolution

        **Outputs**
        - A CSV file containing embeddings. Each row corresponds to an image, with the file name in the first column and embedding vectors in the subsequent columns.
        ]]>
    </help>