Giter Club home page Giter Club logo

Comments (2)

bowenc0221 avatar bowenc0221 commented on August 16, 2024

You can refer to this implementation:

MaskFormer/mask_former/mask_former_model.py

Lines 248 to 304 in da3e60d

def panoptic_inference(self, mask_cls, mask_pred):
scores, labels = F.softmax(mask_cls, dim=-1).max(-1)
mask_pred = mask_pred.sigmoid()
keep = labels.ne(self.sem_seg_head.num_classes) & (scores > self.object_mask_threshold)
cur_scores = scores[keep]
cur_classes = labels[keep]
cur_masks = mask_pred[keep]
cur_mask_cls = mask_cls[keep]
cur_mask_cls = cur_mask_cls[:, :-1]
cur_prob_masks = cur_scores.view(-1, 1, 1) * cur_masks
h, w = cur_masks.shape[-2:]
panoptic_seg = torch.zeros((h, w), dtype=torch.int32, device=cur_masks.device)
segments_info = []
current_segment_id = 0
if cur_masks.shape[0] == 0:
# We didn't detect any mask :(
return panoptic_seg, segments_info
else:
# take argmax
cur_mask_ids = cur_prob_masks.argmax(0)
stuff_memory_list = {}
for k in range(cur_classes.shape[0]):
pred_class = cur_classes[k].item()
isthing = pred_class in self.metadata.thing_dataset_id_to_contiguous_id.values()
mask = cur_mask_ids == k
mask_area = mask.sum().item()
original_area = (cur_masks[k] >= 0.5).sum().item()
if mask_area > 0 and original_area > 0:
if mask_area / original_area < self.overlap_threshold:
continue
# merge stuff regions
if not isthing:
if int(pred_class) in stuff_memory_list.keys():
panoptic_seg[mask] = stuff_memory_list[int(pred_class)]
continue
else:
stuff_memory_list[int(pred_class)] = current_segment_id + 1
current_segment_id += 1
panoptic_seg[mask] = current_segment_id
segments_info.append(
{
"id": current_segment_id,
"isthing": bool(isthing),
"category_id": int(pred_class),
}
)
return panoptic_seg, segments_info

Argmax is used to resolve conflict:

MaskFormer/mask_former/mask_former_model.py

Line 272 in da3e60d

cur_mask_ids = cur_prob_masks.argmax(0)

Pixels without assigned labels are set to VOID which is allowed in panoptic segmentation:

MaskFormer/mask_former/mask_former_model.py

Line 262 in da3e60d

panoptic_seg = torch.zeros((h, w), dtype=torch.int32, device=cur_masks.device)

from maskformer.

Trainingzy avatar Trainingzy commented on August 16, 2024

Cool. I got it. Thanks!

from maskformer.

Related Issues (20)

Recommend Projects

  • React photo React

    A declarative, efficient, and flexible JavaScript library for building user interfaces.

  • Vue.js photo Vue.js

    🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.

  • Typescript photo Typescript

    TypeScript is a superset of JavaScript that compiles to clean JavaScript output.

  • TensorFlow photo TensorFlow

    An Open Source Machine Learning Framework for Everyone

  • Django photo Django

    The Web framework for perfectionists with deadlines.

  • D3 photo D3

    Bring data to life with SVG, Canvas and HTML. 📊📈🎉

Recommend Topics

  • javascript

    JavaScript (JS) is a lightweight interpreted programming language with first-class functions.

  • web

    Some thing interesting about web. New door for the world.

  • server

    A server is a program made to process requests and deliver data to clients.

  • Machine learning

    Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.

  • Game

    Some thing interesting about game, make everyone happy.

Recommend Org

  • Facebook photo Facebook

    We are working to build community through open source technology. NB: members must have two-factor auth.

  • Microsoft photo Microsoft

    Open source projects and samples from Microsoft.

  • Google photo Google

    Google ❤️ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.