Workflows, Best Practice guidance, Tips and Tricks to extract the best from Hammer Missions

Using the Hammer AI Annotation Assistant (to expedite AI Model Creation)

This guide explains how to use the Hammer AI Annotation Assistant to help expedite the creation of your Hammer AI model, which then automatically searches your project for the identified deficiency, assisting you in documenting and quantifying an issue across your entire project model.

https://www.loom.com/share/735b144d830548e9b8fc3e7c53a4e36e

Overview: the workflow in four steps

Prepare the view: lock a facade and orient the 3D model for systematic inspection.
Annotate examples: use the AI Annotation Assistant to help you tag the defect across several images (to create an AI model).
Create the model: submit the annotated examples to build the Hammer AI Model.
Review results: inspect the AI detections & address any detection issues as necessary.

Step 1 — Prepare the model and lock the façade

Start by orientating the 3D model so the elevation you want to inspect faces you squarely. Enable the facade lock so the view maintains orientation as you pan left/right or up/down.

With facade lock on, dragging the left mouse button keeps the perspective fixed rather than rotating the model. The view stability makes it much easier to perform a systematic top‑to‑bottom inspection.

Clear front-facing 3D building model in Hammer Missions with inspection thumbnails visible across the bottom.

Step 2 — Annotate examples to train the Hammer AI

Mark the deficiency that you wish the AI model to identify, e.g. here we chose the beige panels on the facade. Double click the 3D model at the area of interest and then thumbnails for that area will appear in the bottom panel. Pick an image (with the deficiency visible) to work on.

Hammer Missions interface showing the 3D model at left and a selected zoomed facade photo on the right with image thumbnails along the bottom.

How to mark a defect

Click on a thumbnail (it opens in full view, in top right corner of screen), then zoom in to the defect area.
Select the rectangular annotation tool, hold Shift and drag a bounding box around the item.
Give the annotation a relevant & descriptive, name—e.g. beige panel. Note: because we use AI to interpret the annotation name as well as the visual area being annotated, make sure the annotation name is not mis-leading

Hammer Missions interface showing a 'beige panel' annotation and the orange AI assistant button highlighted with a tooltip reading 'Find Similar Annotations'.

After labelling just one example of the deficiency, you can use the Hammer AI annotation assistant (the orange button) to automatically scan the current image, for other occurrences of that same issue. The AI annotation assistant automatically annotates all matches it has detected, so you save time by not having to draw a box around every deficiency manually.

Hammer Missions interface showing many AI-proposed dashed bounding boxes around beige facade panels across a building elevation.

The AI annotation assistant can occasionally miss a deficiency, if that happens, jump to the missed area and draw another bounding box, assign the same label and re-run the assistant for that image. Repeat this annotation process across several images—aim for at least six distinct images containing the specific deficiency you need to identify.

Hammer Missions annotation dialog with the tag 'beige panel' typed and the AI assistant button visible, over a zoomed facade image.

Step 3 — Create the AI deficiency model

Once you have annotated enough examples, open the AI menu in the top right and choose Create. Select the annotation tag you annotated your deficiencies with (for example, beige panel) and continue.

Create AI Models 'Selected Defects' screen showing the beige panel tag with an indicator that the minimum number of images are annotated and a visible Continue button.

The platform verifies if you have enough annotated images (at least 6). Providing you have annotated 6 or more images, you can then move on to name the model, (add a short optional description if you wish) and click Start building. The platform will build an AI model in the background and then run it across every image in the project to find all matching occurrences.

Hammer Missions 'Create AI Models' dialog showing 'Beige panel' model name and a visible 'Start building' button over facade thumbnails.

You will receive an email when the model has finished building. Open the model to inspect the automated detections (you can of course remove any incorrect detections by clicking on their dotted boundary and choosing the delete icon).

Hammer Missions interface with a popup reading 'AI is currently building the model' over facade images and image thumbnails.

Tips and best practices

Choose clear, representative examples: pick images that show the defect from different angles, lighting and distances.
Annotate consistently: use the same label spelling and bounding box style for every example to avoid confusion during training.
Use multiple images: at least six images are recommended—more improves the model’s reliability.
Manually correct misses: AI assistance speeds annotation but may occasionally miss or incorrectly annotate a deficiency. Correct the error and make sure you re-run the AI Quantification option to update your analysis metrics.
Create a model per defect type: treat each recurring deficiency as its own tag and model to keep results precise.

Common pitfalls

Poor image quality or inconsistent framing reduces detection accuracy.
Overlapping or visually similar elements can confuse the model—provide examples that distinguish the defect from lookalikes.
Too few annotated examples will prevent model creation or produce weak results.

Conclusion

Using our platform to build a deficiency specific AI model, quickly turns repetitive visual inspection into a quick, repeatable and scalable process. With consistent annotations and a handful of representative images, you can automate the rapid detection of deficiencies across a whole project and free up time for other key activities.

➡️Next Topic

AI Detection & Quantification

⬅️Previous Topic

3D Model Basics