Horst Obenhaus
Horst Obenhaus
Thanks @Y-T-G ! I wonder: What are those features that are extracted there? I.e. which layer(s) are these embeddings retrieved from? I am getting a `torch.Size([256])` back.
Thanks, do you know what I should / could do to get output from different layers / compositions of layers?
Thanks @glenn-jocher , but these embeddings that are image-wide, not object specific, right? As far as I understand, the answer that @Y-T-G provided (https://github.com/ultralytics/ultralytics/issues/22562#issuecomment-3476997981) is object specific? How do I...
@Y-T-G I do features though for every detected object when I execute my code above? What is happening there (I don't understand why my code works, but the result looks...
Ok, interesting. How do you explain then that I get class specific results (when I loop over results): This is for layer model.12 for example. Is this accidental? It seems...
More specifically I wonder if the combination of YOLO API + my hooks already do some pooling / cropping so I DO get object specificity regardless the layer?
Hey, thanks for your reply and don't worry. I did try the `switch-to-bioio` branch right now and this errors occurs upon opening a .czi (Zeiss) file. This more just to...
Oh beautiful, that did work! Thanks! And its fast and seems to be stable. Cool.
Ok, I was able to help myself here by finding code to extract the longest path after doing `pixel_graph, coordinates = skeleton_to_csgraph(skeleton0)` - it kind of works for my purpose.
> ... skan.summarize is longest-shortest-path Wait, I can do that already in skan?