State Change Temporal Localization and Classification Benchmark

This directory contains baseline implementations for State Change Temporal Localization and Classification Benchmark.

Data Loader

The data loader files are in i3d-resnet50/datasets and are developed and maintained by Siddhant Bansal (https://sid2697.github.io/) at IIIT Hyderabad.

We recommend using the following data loader: https://github.com/EGO4D/hands-and-objects/blob/main/state-change-localization-classification/i3d-resnet50/datasets/StateChangeDetectionAndKeyframeLocalisation.py

I3D ResNet Baseline

The baseline of using I3D ResNet as backbone is in i3d-resnet50/. The baseline is developed and maintained by Siddhant Bansal (https://sid2697.github.io/) at IIIT Hyderabad.

Bi-directional LSTM Baseline

The baseline of using Bi-directional LSTM is in cnn-lstm/. The baseline is developed and maintained by Yifei Huang (https://hyf015.github.io/) at University of Tokyo.

Boundary Matching Network (BMN) Baseline

The baseline of using Boundary Matching Network (BMN) is in bmn/. The baseline is developed and maintained by Raghava Modhugu ([email protected]) at IIIT Hyderabad.

SlowFast + Perceiver Baseline

The baseline of using SlowFast + Perceiver is in slowFast-perceiver/. The baseline is developed and maintained by Abrham Gebreselasie ([email protected]) at Carnegie Mellon University Africa.