Giovanni Cavallin
Giovanni Cavallin
When structuring the configuration, it's useful to keep the training, validation and evaluation configurations separated. In this case, it'd be useful to have a test configuration all alone, letting a...
## Motivation Add Albumentations support to the mmsegmentation framework ## Modification I took inspiration from mmdet and made it compatible with the `mask` key (instead of `masks` and `bboxes`). I've...
When structuring the configuration, it is useful to keep training and validation pipelines separated (in case of heavy augmentations vs original validation dataset). Thanks for your contribution and we appreciate...
Hi, First I'd like to congrat you about your great work! I'd like to ask whether a single precision training has been taken into consideration for the future, or if...
After some updates (ppa not supporting python 3.6 installations anymore, and I think many changes in the `tensorflow/tensorflow:custom-op-ubuntu16` base image), I was not able to build the image anymore. I...
## Motivation The function `mmrotate.structures.bbox.bbox_overlaps.rbbox_overlaps does not accept eps as parameter, so it shouldn't be passed here. I think it should be passed until here because: - At this point,...
Hi there, thank you for your groundbreaking work. I'm trying to develop the rotation equivariant version of [this](https://openaccess.thecvf.com/content/ICCV2023/papers/Li_Large_Selective_Kernel_Network_for_Remote_Sensing_Object_Detection_ICCV_2023_paper.pdf) neural network. I succesfully replaced every layer of the network, but there...
## Motivation The switch pipeline was not compatible with MultImageMixDataset: it only replaced the `loading_pipeline` of the outer (`MultiImageMixDataset`) dataset. With this modification, the loading pipeline is set to the...
## Motivation Bugfixing CopyPaste: indices multiplication was creating a `results['img']` `int64` dtype, now it's casted to the `src_img` dtype. ## Checklist 1. Pre-commit or other linting tools are used to...