Dian Chen

Modifications to ResNet-50

1

Hi, I loaded the r50_1x_sk0 checkpoint for ResNet-50, and it seems that the network architecture isn't exactly the original one in [ResNet paper](https://arxiv.org/pdf/1512.03385.pdf). However, I didn't see this is mentioned...

Loss goes nan after 14 epochs

1

Hi, thank you for releasing such a wonderful work. I tried to replicate the results using the following command: ``` python -m torch.distributed.launch --nproc_per_node 8 main_simmim.py --cfg configs/swin_base__100ep/simmim_pretrain__swin_base__img192_window6__100ep.yaml --data-path /mnt/fsx/datasets/imagenet/train...

Data augmentation with segmentation map/queries

1

Hi! I noticed that in your `petrv2_BEVseg.py` config file there's `GlobalRotScaleTransImage` which applies rotation and scaling to the bev space and correspondingly 3d ground truth boxes (which makes total sense...

Support of customized models

1

Hi! Does pytorch-cpp support any conversion from customized models built in pytorch, or does it only support standard models (e.g. vgg, res-net) and your own segmentation models? Big thanks!

Hi! Thank you for releasing this great work. May I ask in you NuScenes leaderboard reasult `BEVDepth-pure`, which checkpoint of your convnext backbone is used (there are many options on...

Hi @yinchimaoliang @ZRandomize ! Thank you for releasing this wonderful work. I have some questions for the BEV features and voxel pooling: 1. What's the coordinate mapping between the ego...

How to render instance id segmentation masks

2

## ❓ Questions on how to use PyTorch3D Hi! Thank you for maintaining this wonderful repo! Let's say I have meshes of objects, with their instance ids (or semantic category...

Hi! Thank you for releasing this wonderful work. I have a question on how RFR operates: Since the dependencies of F_t^i are the items to its LEFT and TOP (like...

About feature enhancer architecture

4

Hi, I'm looking at `groundingdino/models/GroudingDINO/transformer.py`, from [line 545 - 593](https://github.com/IDEA-Research/GroundingDINO/blob/60d796825e1266e56f7e4e9e00e88de662b67bd3/groundingdino/models/GroundingDINO/transformer.py#L545-L593) it looks like the order of modules is `Bi-Direction MHA (text->image, image->text)` -> `text self-attention` and `image deformable self-attention`, which...

Question on training label

1

Hi! Thank you for release this interesting work. Just to check my understanding: during few-shot training, is the label made by pseudo labeling with the _original_ CLIP feature and classification...