Skip to content

This CL is a modified version of the TAPNext model that can handle video segmentation, alongside other modifications, such as the kubric data reader. #1163

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
wants to merge 1 commit into
base: main
Choose a base branch
from
Open

This CL is a modified version of the TAPNext model that can handle video segmentation, alongside other modifications, such as the kubric data reader. #1163

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
5 changes: 4 additions & 1 deletion scenic/projects/baselines/segment_anything/modeling/mask_decoder.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -58,7 +58,10 @@ def setup(self):
num_layers=self.iou_head_depth,
name='iou_prediction_head')

self.transformer = transformer.TwoWayTransformer(name='transformer')
self.transformer = transformer.TwoWayTransformer(
name='transformer',
embedding_dim=self.transformer_dim,
)

def predict_masks(
self, image_embeddings, image_pe,
Expand Down
5 changes: 4 additions & 1 deletion scenic/projects/baselines/segment_anything/modeling/prompt_encoder.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -32,8 +32,8 @@ class PromptEncoder(nn.Module):
"""Sam prompt encoder for points and boxes."""

embed_dim: int = 256
image_embedding_size: Tuple[int, int] = (1024 // 16, 1024 // 16)
input_image_size: Tuple[int, int] = (1024, 1024)
image_embedding_size: Tuple[int, int] = (1024 // 16, 1024 // 16)
num_point_embeddings: int = 4 # pos/neg point + 2 box corners
mask_in_chans: int = 16

Expand Down Expand Up @@ -114,6 +114,7 @@ def _embed_boxes(self, boxes, image_size=None):
return corner_embedding

def _embed_masks(self, masks):
print(masks.shape)
mask_embedding = self.mask_downscaling(masks)
return mask_embedding

Expand Down Expand Up @@ -156,6 +157,7 @@ def __call__(
sparse_embeddings = jnp.concatenate(
[sparse_embeddings, box_embeddings], axis=1)
if masks is not None:
print(masks.shape)
dense_embeddings = self._embed_masks(masks)
else:
if image_embedding_size is None:
Expand All @@ -165,6 +167,7 @@ def __call__(
(num_prompts, image_embedding_size[0],
image_embedding_size[1], self.embed_dim,)
)

return sparse_embeddings, dense_embeddings


Expand Down