TopicFM
TopicFM copied to clipboard
finetuning with correspondences
Can the model be fine-tuned with images and their corresponding labels, without using a depth map?
What do you mean by "corresponding labels" here? semantic segmentation labels?