pranavdurai10/rtdetr-pytorch

This repository provides a PyTorch implementation of RT-DeTR, a state-of-the-art Realtime Detection Transformer for object detection tasks.

Hello, World! 👋

About Me

I am Pranav Durai, a Sr. Computer Vision Engineer at OpenCV. I'm a passionate Quantum AI Researcher, and currently, I serve as a Research Fellow at Stanford University - School of Medicine, focusing on pre-clinical medical imaging and segmentation. My love for nature led me to explore its wonders, traveling thousands of kilometers on my motorcycle in search of self-discovery. As an enthusiastic Apple fan, I have been devotedly following the company since 2010. My mission is to contribute to humanity's sustainable future through my work.

DARE MIGHTY THINGS 🚀

rtdetr-pytorch's Issues

Review and build: RTDeTR decoder

RTDeTR decoder implementation detr_layers/decoder.py needs to be validated with actual architecture.

Please feel free to submit a PR!

performance?

performance with this code?

Review: DWConv

DWConv implementation detr_layers/dwconv.py needs to be validated with actual architecture.

Please feel free to submit a PR!

Review: AIFI layer

AIFI layer implementation detr_layers/aifi.py needs to be validated with actual architecture.

Please feel free to submit a PR!

Review: HGBlock

HGBlock implementation detr_layers/hgblock.py needs to be validated with actual architecture.

Please feel free to submit a PR!

Dino: Re-build transformer head module for L and X RTDeTR models

Dino head needs to be built for both L and X models in models directory.

Please feel free to submit a PR!

Review: Concat layer

Concat layer implementation detr_layers/concat.py needs to be validated with actual architecture.

Please feel free to submit a PR!

Review: Conv layer

Conv layer implementation detr_layers/conv.py needs to be validated with actual architecture.

Please feel free to submit a PR!

Complete model

Thanks for sharing! I'd like to do some research based on this library. Compared with the original paper, what modules do we need to complete this model, in addition to the ones mentioned in the issue?
Or do you have further more complete work now？

Review: HGStem

HGStem implementation detr_layers/hgstem.py needs to be validated with actual architecture.

Please feel free to submit a PR!

Review: RepC3 layer

RepC3 layer implementation detr_layers/repc3.py needs to be validated with actual architecture.

Please feel free to submit a PR!

请问该模型实现的训练结果与paddle原版对比如何

又没有对cocomin或者coco进行对比实验，训练结果如何，能达到paddle一样的准确率吗？