This project aims to build a new CODEBASE for Scene Graph Generation (SGG), and it is also a Pytorch
implementation of the paper "Unbiased Scene Graph Generation from Biased Training". It is built on top of
the well-known maskrcnn-benchmark and defines relationship prediction as an additional roi_head. Moreover,
I included all the exsiting metrics: R@K, mR@K, ngR@K, zR@K, to benchmark the SGG.
This project provides a strong single-stage baseline for Long-Tailed Classification (under ImageNet-LT,
Long-Tailed CIFAR-10/-100 datasets), Detection, and Instance Segmentation (under LVIS dataset). It is also
a PyTorch implementation of the NeurIPS 2020 paper Long-Tailed Classification by Keeping the Good and
Removing the Bad Momentum Causal Effect. This project can be easily generalized to other tasks with unbalanced datasets.
An open-source visual question answering (VQA) CODEBASE built on top of the bottom-up-attention-vqa. It
integrates several popular VQA papers published in 2018, which includes: bottom-up top-down, bilinear
attention network, learning to count, learning conditioned graph structures, intra- and inter-modality
attention.