XMem2
XMem2 copied to clipboard
max810
A tool for efficient semi-supervised video object segmentation (great results with minimal manual labor) and a dataset for benchmarking
When I run your code, it constantly prompts "Finish saving the results, 15 masks and 0 overlays left." and no further inference is made
On our PC on the same video for one time propagating over all frames of the same video Xmem uses just 3 GB GPU memory while your structure uses 9GB...
Hi. Thanks for the great work. I've tried to use the inference script to propagate the annotations from multiple key frames (e.g., 5) to all video frames. However, I find...
Added a text box to specify the current desired frame to have an alternative to the slider. Changed defaults to the values Lap recommended. We may something more flexible than...
Traceback (most recent call last): File "C:\Users\phili\Documents\XMem2-main\XMem2-main\inference\interact\gui.py", line 1055, in on_mouse_release self.vis_map, self.vis_alpha = interaction.push_point(ex, ey, ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "C:\Users\phili\Documents\XMem2-main\XMem2-main\inference\interact\interaction.py", line 225, in push_point self.obj_mask = self.controller.interact(self.image.unsqueeze(0), x, y, not neg)...
Hi! Great job with the project. I have a question about the visualization images. Is it possible not to generate them to save some processing time? If the answer is...
Thank you for the project, great tool you created. I found a Pretty small bug regarding file endings. If I try to load a file sequence with the fileending *.jpeg...
I experienced crashes, and found the problem. In the file brs_functors line 75 the last "np.float" should be "np.cfloat" ``` f_grad = opt_params.grad.cpu().numpy().ravel().astype(np.float) #
