LinasKo

It's alright; we've all been through that :wink: Thank you very much for your contribution. I'll make sure it gets verified and merged soon!

@SkalskiP, Ready to review. I've closed prior PR, and will include InferenceSlicer integration in a separate one. Tests on 2 images, benchmarks. no videos or segmentation https://colab.research.google.com/drive/1v0MPlG1tQctX5-koh0l6h1NcB6eWJ_YY?usp=sharing I haven't tested...

Hey @SkalskiP, I pushed an update; here's some questions: 1. Removed `non_max_merge` altogether - it wasn't used. 1. There's now tests for `merge_object_detection_pair` and `box_non_max_merge`, but not for `batch_box_non_max_merge`. Would...

Thanks, that's helpful :) * I've now realized that for the method we planned to move to utils, annotating with `"Detections"`. Is not enough. It needs to know about the...

Added `np.NDArray` types for functions affected by this change. I believe I've addressed the change requests.

Hi @SkalskiP, I've reviewed the [HF transformers](https://github.com/huggingface/transformers/blob/fed27ffc7ec62837dca9bbfc83442eb3678ee026/src/transformers/models/detr/image_processing_detr.py) code - #1054 didn't add support for instance and panoptic segmentation. The result dicts from HF are much different in those cases. I...

@SkalskiP The missing part is where I suggested raising `NoImplementedError` which passing instance and panoptic seg results would trigger.

# Affected code The good news is - the error mostly appears when `Detections.merge` is called on results from _unusual models_ - where `confidence` is missing, where `data` is set,...

# Mitigation Options **TL;DR: I think Option 3 is the least-bad solution. Adds complexity in exactly one location, keeps design and architecture intact, stops errors, prevents future worries. I also...

We've now implemented option 3 in https://github.com/roboflow/supervision/pull/1177 Ideally, I'd like to see **an attempt at** trimming down `Detections.empty` to just `xyxy` before closing this. Would that cause issues? Would that...