Official implementation of STREAMER, a self-supervised hierarchical event segmentation and representation learning
Project description
STREAMER
The official PyTorch implementation of our NeurIPS'23 paper STREAMER: Streaming Representation Learning and Event Segmentation in a Hierarchical Manner
Overview
Documentation
Checkout the documentation of STREAMER modules to learn more details about how to use our codebase.
Installation
pip install streamer-torch # with pip from PyPI
pip install git+'https://github.com/ramyamounir/streamer-torch' # with GitHub
Inference
Note: Pretrained weights are coming soon..
from streamer.models.inference_model import InferenceModel
model = InferenceModel(checkpoint='to/checkpoint/path/')
result = model(filename='to/video/file/path')
Training
In order to perform training with streamer:
- Use the Dataset README.md to preprocess datasets for streaming loading and evaluation.
- Use the provided training script to train on multiple gpus (i.e., or multi-node).
- The script compare.py can be used to evaluate the model's prediction using Hierarchical Level Reduction.
Bash scripts with CLI arguments are provided in helper_scripts
Citing STREAMER
If you find our approaches useful in your research, please consider citing:
@inproceedings{mounir2023streamer,
title={STREAMER: Streaming Representation Learning and Event Segmentation in a Hierarchical Manner},
author={Mounir, Ramy and Vijayaraghavan, Sujal and Sarkar, Sudeep},
booktitle={Thirty-seventh Conference on Neural Information Processing Systems},
year={2023}
}
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
streamer-torch-0.0.3.tar.gz
(27.1 kB
view hashes)
Built Distribution
Close
Hashes for streamer_torch-0.0.3-py3-none-any.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | 88c9c6b24ada066a8c9e742b29b71337fe6c8ee51d22a80019afa72d1c561ef2 |
|
MD5 | 665ce195e47c334d3e5d9d0ba7c3dd5a |
|
BLAKE2b-256 | d83b0cf42d6f6a05b9f3b0f359a7b9668d133468d367b02577438ce1b07e4aea |