#designbase

Sets of buildings are generated that have spaces (residential, Airbnb, and hotel) on high-traffic tourist routes.

Interventions are located with the aim of creating the network.

These complexes can be understood as franchises of the resort.

Moodboard 01.

Moodboard 02.

Connection.

Step 1: Modify the Chrome shortcut's target path and launch the browser using the modified shortcut. Record the PID (Process ID) shown in the pop-up window.

Step 2: In RenderDoc, locate the Chrome GPU rendering process using the PID and inject RenderDoc as a monitor into this process.

Step 3: Open Google Earth in the browser and navigate to the area you want to capture. After the model has fully loaded, click the Capture button in RenderDoc. During the capture, keep moving or rotating the view to ensure the GPU is actively rendering, allowing RenderDoc to successfully intercept the data. Once finished, export the .rdc file.

Step 4: Install the corresponding RenderDoc import plugin in Blender, and load the .rdc file to retrieve the captured 3D model.

The final model is combined with the basic model we had without detail to highlight the groups of buildings that will make up the resort's sites.

Image:

Video:

3D:

Original:

Image:

Video:

3D:

Original space:

The shared steps between the two methods：

Step 1: Extract all frames from the input video.

Step 2: Extract keypoint features from each frame, then match features between adjacent frames. Based on these matches, estimate camera motion to generate a sparse point cloud.

Step 3: Estimate depth relationships based on the sparse point cloud to generate a dense point cloud.

The different steps between the two methods：

Gauss splatting：Each point in the dense point cloud is fitted with a 3D Gaussian distribution, where parameters include position, orientation (anisotropy), scale, color, and opacity.