About 3.4 Basis-sharing conversion。

[256-7421142]

“Finally, we retrieve the object boundary by using the outer edge of the union of the object’s local contours.”
Where is the corresponding code for this part?（I have got the mask_ap）

[tding1]

Thanks for the question, will update that part later :)

[256-7421142]

“Finally, we retrieve the object boundary by using the outer edge of the union of the object’s local contours.” Where is the corresponding code for this part?（I have got the mask_ap）
Thanks for the reply， I wonder why there will be errors in the encoding or convert stage, and why the F-measure is very low in the convert stage?

[256-7421142]

Thanks for the question, will update that part later :)

Thanks for the question, will update that part later :)
It would be great if the code for this part of the instance segmentation could be published. Because I encountered some difficulties applying Adacontour in my network to train the custom dataset.

[256-7421142]

Thanks for the question, will update that part later :)

Excuse me, I would like to know how to visualise U, looking forward to your reply!

[256-7421142]

Thanks for the question, will update that part later :)

Sorry，the rest code I refer to this，could you upload this part of the code earlier?

[tding1]

its already here:

AdaContour/libs/S3_convert.py

Lines 45 to 54 in a895756

	meta = out["meta" + str(i)]
	contour_center = meta["contour_center"][()]
	r = meta["r"][()]
	seg_gt = to_np(self.label_all[i]["seg_mask"][0])
	U = self.U[:, : self.cfg.dim]
	U_t = (U.T).copy()
	c = r @ U / self.cfg.max_dist
	r_ap = c @ U_t * self.cfg.max_dist

Please go through the README file for running the entire pipeline on the demo dataset.

[256-7421142]

its already here:

AdaContour/libs/S3_convert.py

Lines 45 to 54 in a895756

	meta = out["meta" + str(i)]
	contour_center = meta["contour_center"][()]
	r = meta["r"][()]
	seg_gt = to_np(self.label_all[i]["seg_mask"][0])
	U = self.U[:, : self.cfg.dim]
	U_t = (U.T).copy()
	c = r @ U / self.cfg.max_dist
	r_ap = c @ U_t * self.cfg.max_dist

Please go through the README file for running the entire pipeline on the demo dataset.
I have run the entire pipline on sbd training set and got the display_approx, but the contours are separated local contours, not the outer edge of the union of the object’s local contours. What's going wrong?
I send the first two display_approx files to your email already!

[256-7421142]

its already here:

AdaContour/libs/S3_convert.py

Lines 45 to 54 in a895756

	meta = out["meta" + str(i)]
	contour_center = meta["contour_center"][()]
	r = meta["r"][()]
	seg_gt = to_np(self.label_all[i]["seg_mask"][0])
	U = self.U[:, : self.cfg.dim]
	U_t = (U.T).copy()
	c = r @ U / self.cfg.max_dist
	r_ap = c @ U_t * self.cfg.max_dist

Please go through the README file for running the entire pipeline on the demo dataset.

I am not sure which line code can implement the function described in this sentence
“Finally, we retrieve the object boundary by using the outer edge of the union of the object’s local contours.”