No decoder

No mask generator

No mask generator

No mask token

No mask token

No 'prediction head' for downstream tasks.

No prediction head for downstream tasks, define your own head

Decay the learning rate with half-cycle cosine after warmup

Use accelerator to all-reduce and compute mean.

Helper function for model output during inference.

flow_utils.colorTest(): display an example of image showing the color encoding scheme

Views x with the dimensions from dim_begin to dim_end folded.

Copies the code directory into the path specified using rsync. It will use a .gitignore file to exclude files in rsync. We preserve modification

Compute the edge mask of a depth map. The edge is defined as the pixels whose neighbors have a large difference in depth. Args:

Args: - depthmap (HxW array): - camera_intrinsics: a 3x3 matrix - camera_pose: a 4x3 or 4x4 cam2world matrix Returns:

DINOv2 ViT-B/14 model (optionally) pretrained on the LVD-142M dataset.

DINOv2 ViT-B/14 model with registers (optionally) pretrained on the LVD-142M dataset.

DINOv2 ViT-g/14 model (optionally) pretrained on the LVD-142M dataset.

DINOv2 ViT-g/14 model with registers (optionally) pretrained on the LVD-142M dataset.

DINOv2 ViT-L/14 model (optionally) pretrained on the LVD-142M dataset.

DINOv2 ViT-S/14 model (optionally) pretrained on the LVD-142M dataset.

DINOv2 ViT-S/14 model with registers (optionally) pretrained on the LVD-142M dataset.

Downsamples a list of image names by keeping every `downsample_factor`-th image. Args: image_names (list of str): List of image

transform a python obj or cmd into a proper filename. - \1 gets replaced by slash '/' - \2 gets replaced by comma ','

Fix random seeds.

Args: query: Tensor of shape (B, N, C), input query key: Tensor of shape (B, M, C), input key value: Tens

Args: query: Tensor of shape (B, N, C), input query key: Tensor of shape (B, M, C), input key value: Tens

input: * tokens: batch_size x nheads x ntokens x dim * positions: batch_size x ntokens x 2 (y and x posit