[[["容易理解","easyToUnderstand","thumb-up"],["確實解決了我的問題","solvedMyProblem","thumb-up"],["其他","otherUp","thumb-up"]],[["難以理解","hardToUnderstand","thumb-down"],["資訊或程式碼範例有誤","incorrectInformationOrSampleCode","thumb-down"],["缺少我需要的資訊/範例","missingTheInformationSamplesINeed","thumb-down"],["翻譯問題","translationIssue","thumb-down"],["其他","otherDown","thumb-down"]],["上次更新時間:2025-09-04 (世界標準時間)。"],[],[],null,["# Configure visualization settings for AutoML image classification\n\n| **Preview**\n|\n|\n| This feature is subject to the \"Pre-GA Offerings Terms\" in the General Service Terms section\n| of the [Service Specific Terms](/terms/service-terms#1).\n|\n| Pre-GA features are available \"as is\" and might have limited support.\n|\n| For more information, see the\n| [launch stage descriptions](/products#product-launch-stages).\n\nVertex Explainable AI provides built-in visualization capabilities for your image data.\nYou can configure visualizations for AutoML image classification\nmodels.\n\nWhen you request an explanation on an image classification model, you get the\npredicted class along with an image overlay showing which pixels\n(integrated gradients) or regions (integrated gradients or XRAI) contributed to\nthe prediction.\n\nThe following images show visualizations on a husky image. The left\nvisualization uses the integrated gradients method and highlights areas of\npositive attribution. The right visualization uses an XRAI method with a color\ngradient indicating areas of lesser (blue) and greater (yellow) influence in\nmaking a positive prediction. \n\nThe type of data you're working with can influence whether you use an\nintegrated gradients or XRAI approach to visualizing your explanations.\n\n- XRAI tends to be better with natural images and provides a better high-level summary of insights, like showing that positive attribution is related to the shape of a dog's face.\n- Integrated gradients (IG) tends to provide details at the pixel level and is useful for uncovering more granular attributions.\n\nLearn more about the attribution methods in the Vertex Explainable AI\n[Overview page](/vertex-ai/docs/explainable-ai/overview).\n\nGetting started\n---------------\n\nConfigure visualization when you\n[train an AutoML model that supports Vertex Explainable AI](/vertex-ai/docs/training/automl-console)\nand\n[enable explanations when you deploy the model](/vertex-ai/docs/predictions/deploy-model-console).\n\nVisualization options\n---------------------\n\nThe default and recommended settings depend on the attribution method\n(integrated gradients or XRAI). The following list describes configuration\noptions and how you might use them. For a full list of options, see the\n[API reference for the `Visualization` message](/vertex-ai/docs/reference/rest/v1/ExplanationSpec#visualization).\n\n- `type`: The type of visualization used: `OUTLINES` or `PIXELS`. Only specify\n this field if you are using integrated gradients; you can't specify it if you\n are using XRAI.\n\n For integrated gradients, the field defaults to `OUTLINES`, which shows\n regions of attribution. To show per-pixel attribution, set the field to\n `PIXELS`.\n- `polarity`: The directionality of the highlighted attributions. `positive` is\n set by default, which highlights areas with the highest positive\n attributions. This means highlighting pixels that\n were most influential to the model's positive prediction.\n Setting polarity to `negative` highlights areas that lead the model to not\n predicting the positive class. Using a negative polarity can be useful for\n debugging your model by identifying false negative regions. You can also set\n polarity to `both` which shows positive and negative attributions.\n\n- `clip_percent_upperbound`: Excludes attributions above the specified\n percentile\n from the highlighted areas. Using the clip parameters together can be useful\n for filtering out noise and making it easier to see areas of strong\n attribution.\n\n- `clip_percent_lowerbound`: Excludes attributions below the specified\n percentile\n from the highlighted areas.\n\n- `color_map`: The color scheme used for the highlighted areas. Default is\n `pink_green` for integrated gradients, which shows positive attributions in\n green and negative in pink. For XRAI visualizations, the color map is a\n gradient. The XRAI default is `viridis` which highlights the most influential\n regions in yellow and the least influential in blue.\n\n For a full list of possible values, see the [API reference for the\n `Visualization` message](/vertex-ai/docs/reference/rest/v1/ExplanationSpec#visualization).\n- `overlay_type`: How the original image is displayed in the visualization.\n Adjusting the overlay can help increase visual clarity if the original image\n makes it difficult to view the visualization.\n\n For a full list of possible values, see the [API reference for the\n `Visualization` message](/vertex-ai/docs/reference/rest/v1/ExplanationSpec#visualization).\n\nExample configurations\n----------------------\n\nTo get started, here are sample `Visualization` configurations that you can use\nas a starting point and images that show a range of settings applied.\n\n### Integrated gradients\n\nFor integrated gradients, you may need to adjust the clip values if the\nattribution areas are too noisy. \n\n visualization: {\n \"type\": \"OUTLINES\",\n \"polarity\": \"positive\",\n \"clip_percent_lowerbound\": 70,\n \"clip_percent_upperbound\": 99.9,\n \"color_map\": \"pink_green\",\n \"overlay_type\": \"grayscale\"\n }\n\nThe following are two visualizations using both the `outlines` and `pixels`\ntypes. The columns labeled \"Highly predictive only,\" \"Moderately predictive,\"\nand \"Almost all\" are examples of clipping at different levels that can help\nfocus your visualization.\n\n### XRAI\n\nFor XRAI visualizations, we recommend starting with no clip values for\nXRAI because the overlay uses a gradient to show areas of high and low\nattribution. \n\n visualization: {\n \"clip_percent_lowerbound\": 0,\n \"clip_percent_upperbound\": 100,\n \"color_map\": \"viridis\",\n \"overlay_type\": \"grayscale\"\n }\n\nThe following image is an XRAI visualization using the default viridis color map\nand a range of overlay types. The areas in yellow indicate the most\ninfluential regions that contributed positively to the prediction.\n\nWhat's next\n-----------\n\n- Use [Getting explanations](/vertex-ai/docs/explainable-ai/getting-explanations) to get predictions with explanations from your model.\n- For details about improving Vertex Explainable AI results for AutoML image classification models, read [Improving explanations](/vertex-ai/docs/explainable-ai/improve-explanations-automl-icn)."]]