Not displaying class labels/captions in BoundingBoxes2D images

[leander]

Hi all! I’m logging some images to my dashboard using this code:

box_data = [{
  "position": {
  "minX": xywh[0] - 3,  # ignore the weird  -3/+3 hardcoded offsets
  "minY": xywh[1] - 3,
  "maxX": xywh[0] + 3,
  "maxY": xywh[1] + 3},
  "class_id": int(cls),
  "scores": {
  "class_score": conf},
  "domain": "pixel"} for *xywh, conf, cls in pred.tolist()]
boxes = {"predictions": {"box_data": box_data, "class_labels": names}}  # inference-space
self.bbox_media_panel_images.append(wandb.Image(im, boxes=boxes, caption=path.name))

My images are filled with tiny cells and I want to have the cleanest visualization of my bounding boxes, without any captions/class labels. Therefore I’m purposefully not filling in the box_caption arg, but the visualization seems to default back to the class_id label, which is mandatory. I then still get images where the labels obscure other cells:

Any way of just visualizing the bounding boxes without captions or labels?

[leander]

Bump, this is getting buried

[bill-morrisson]

Hi Leander,

Thank you for your patience and for contacting us about this.
If I understand correctly what you are trying to do and from your code and visualization, I see that you have assigned a value to the class_id  label but I do not see any box_caption label. I suppose that can be the reason?
Please add the option and let me know if there is still an issue?

[bill-morrisson]

Hi Leander,

Checking out if you tried my suggestion?

[leander]

Hi Bill! Sorry for the slow response. I filled in box_caption with an empty string "", but this just makes the visualization show empty label boxes of arguably the same size as when I had filled them with the class_ids:

box_caption=None throws an error (it has to be a string).