Closed GuangtaoLyu closed 2 months ago
We didn't use (and didn't have) any training data with obstacles for training. As our method does not require training a new model, we manually create random obstacles for testing, represented as signed distance functions. For simplicity, we only use sphere-shaped obstacles because creating a good SDF from an arbitrary scene scan is a challenge on its own.
hello, thank you for your great work. How can i find some related papers and datasets about avoiding obstacles? I only find some words in your paper. Can you share more details?