Implementation of TimeSformer, a pure attention-based solution for video classification

Related tags

Deep Learningdeep-learningtransformersartificial-intelligenceattention-mechanismvideo-classification
Overview

TimeSformer - Pytorch

Implementation of TimeSformer, a pure and simple attention-based solution for reaching SOTA on video classification. This repository will only house the best performing variant, 'Divided Space-Time Attention', which is nothing more than attention along the time axis before the spatial.

Install

$ pip install timesformer-pytorch

Usage

import torch
from timesformer_pytorch import TimeSformer

model = TimeSformer(
    dim = 512,
    image_size = 224,
    patch_size = 16,
    num_frames = 8,
    num_classes = 10,
    depth = 12,
    heads = 8,
    dim_head =  64,
    attn_dropout = 0.1,
    ff_dropout = 0.1
)

video = torch.randn(2, 8, 3, 224, 224) # (batch x frames x channels x height x width)
pred = model(video) # (2, 10)

Citations

@misc{bertasius2021spacetime,
    title   = {Is Space-Time Attention All You Need for Video Understanding?}, 
    author  = {Gedas Bertasius and Heng Wang and Lorenzo Torresani},
    year    = {2021},
    eprint  = {2102.05095},
    archivePrefix = {arXiv},
    primaryClass = {cs.CV}
}
Comments
  • How to deal with varying length video? Thanks

    How to deal with varying length video? Thanks

    Dear all, I am wondering if TimeSformer can handle different videos with diverse lengths? Is it possible to use mask as the original Transformer? Any ideas, thanks a lot.

    opened by junyongyou 2
  • fix runtime error in SpaceTime Attention

    fix runtime error in SpaceTime Attention

    There is a shape mismatch error in Attention. When we splice out the classification token from the first token of each sequence in q, k and v, the shape becomes (batch_size * num_heads, num_frames * num_patches - 1, head_dim). Then we try to reshape the tensor by taking out a factor of num_frames or num_patches (depending on whether it is space or time attention) from dimension 1. That doesn't work because we subtracted out the classification token.

    I found that performing the rearrange operation before splicing the token fixes the issue.

    I recreate the problem and illustrate the solution in this notebook: https://colab.research.google.com/drive/1lHFcn_vgSDJNSqxHy7rtqhMVxe0nUCMS?usp=sharing.

    By the way, thank you to @lucidrains; all of your implementations on attention-based models are helping me more than you know.

    opened by adam-mehdi 1
  • Update timesformer_pytorch.py

    Update timesformer_pytorch.py

    fixing issue for scaling

    File "/home/aarti9/.local/lib/python3.6/site-packages/timesformer_pytorch/timesformer_pytorch.py", line 82, in forward q *= self.scale

    RuntimeError: Output 0 of ViewBackward is a view and is being modified inplace. This view is an output of a function that returns multiple views. Inplace operators on such views is forbidden. You should replace the inplace operation by an out-of-place one.

    opened by aarti9 0
  • Fine-tune with new datasets

    Fine-tune with new datasets

    Thank you so much for your great effort. I can predict the images using the given .py files. But, I couldn't find train.py files, so how to fine-tune the network with new datasets? where should i define the image samples of the new dataset ?

    opened by Jeba-create 0
  • problem in timesformer_pytorch.py

    problem in timesformer_pytorch.py

    start from line 182 video = rearrange(video, 'b f c (h p1) (w p2) -> b (f h w) (p1 p2 c)', p1 = p, p2 = p) i think this should be video = rearrange(video, 'b f c (hp p1) (wp p2) -> b (f hp wp) (p1 p2 c)', p1 = p, p2 = p)

    opened by Weizhongjin 2
  • Imagenet Pretrained Weights

    Imagenet Pretrained Weights

    Thanks for the work! In their paper they say For all our experiments, we adopt the “Base” ViT model architecture (Dosovitskiy et al., 2020) pretrained on ImageNet.

    I know that you said the official weights trained on kinetics and such are not officially released yet. However, I am not interested in those but am actually in need of the initial weights of the network just based on ViT Imagenet pretraining. I need to train this implementation of yours starting from those. From what it looks like, you don't have weights for this implementation that come from imagenet pretraining, do you?

    opened by RaivoKoot 5
Owner
Phil Wang
Working with Attention. It's all we need.
Phil Wang
GitHub Repository
TensorFlow implementation of the paper "Hierarchical Attention Networks for Document Classification"

Hierarchical Attention Networks for Document Classification This is an implementation of the paper Hierarchical Attention Networks for Document Classi

Quoc-Tuan Truong 83 Dec 05, 2022
Create animations for the optimization trajectory of neural nets

Animating the Optimization Trajectory of Neural Nets loss-landscape-anim lets you create animated optimization path in a 2D slice of the loss landscap

Logan Yang 81 Dec 25, 2022
A New Open-Source Off-road Environment for Benchmark Generalization of Autonomous Driving

A New Open-Source Off-road Environment for Benchmark Generalization of Autonomous Driving Isaac Han, Dong-Hyeok Park, and Kyung-Joong Kim IEEE Access

13 Dec 27, 2022
3.8% and 18.3% on CIFAR-10 and CIFAR-100

Wide Residual Networks This code was used for experiments with Wide Residual Networks (BMVC 2016) http://arxiv.org/abs/1605.07146 by Sergey Zagoruyko

Sergey Zagoruyko 1.2k Dec 29, 2022
SeqFormer: a Frustratingly Simple Model for Video Instance Segmentation

SeqFormer: a Frustratingly Simple Model for Video Instance Segmentation SeqFormer SeqFormer: a Frustratingly Simple Model for Video Instance Segmentat

Junfeng Wu 298 Dec 22, 2022
Code of Classification Saliency-Based Rule for Visible and Infrared Image Fusion

CSF Code of Classification Saliency-Based Rule for Visible and Infrared Image Fusion Tips: For testing: CUDA_VISIBLE_DEVICES=0 python main.py For trai

Han Xu 14 Oct 31, 2022
An attempt at the implementation of Glom, Geoffrey Hinton's new idea that integrates neural fields, predictive coding, top-down-bottom-up, and attention (consensus between columns)

GLOM - Pytorch (wip) An attempt at the implementation of Glom, Geoffrey Hinton's new idea that integrates neural fields, predictive coding,

Phil Wang 173 Dec 14, 2022
Discord bot for notifying on github events

Git-Observer Discord bot for notifying on github events ⚠️ This bot is meant to write messages to only one channel (implementing this for multiple pro

ilu_vatar_ 0 Apr 19, 2022
Official implementation for "Low-light Image Enhancement via Breaking Down the Darkness"

Low-light Image Enhancement via Breaking Down the Darkness by Qiming Hu, Xiaojie Guo. 1. Dependencies Python3 PyTorch=1.0 OpenCV-Python, TensorboardX

Qiming Hu 30 Jan 01, 2023
PyTorch implementation of neural style randomization for data augmentation

README Augment training images for deep neural networks by randomizing their visual style, as described in our paper: https://arxiv.org/abs/1809.05375

84 Nov 23, 2022
A ssl analyzer which could analyzer target domain's certificate.

ssl_analyzer A ssl analyzer which could analyzer target domain's certificate. Analyze the domain name ssl certificate information according to the inp

vincent 17 Dec 12, 2022
Interactive dimensionality reduction for large datasets

BlosSOM 🌼 BlosSOM is a graphical environment for running semi-supervised dimensionality reduction with EmbedSOM. You can use it to explore multidimen

19 Dec 14, 2022
Simple reference implementation of GraphSAGE.

Reference PyTorch GraphSAGE Implementation Author: William L. Hamilton Basic reference PyTorch implementation of GraphSAGE. This reference implementat

William L Hamilton 861 Jan 06, 2023
[NeurIPS 2021] ORL: Unsupervised Object-Level Representation Learning from Scene Images

Unsupervised Object-Level Representation Learning from Scene Images This repository contains the official PyTorch implementation of the ORL algorithm

Jiahao Xie 55 Dec 03, 2022
Arabic Car License Recognition. A solution to the kaggle competition Machathon 3.0.

Transformers Arabic licence plate recognition 🚗 Solution to the kaggle competition Machathon 3.0. Ranked in the top 6️⃣ at the final evaluation phase

Noran Hany 17 Dec 04, 2022
Time-stretch audio clips quickly with PyTorch (CUDA supported)! Additional utilities for searching efficient transformations are included.

Time-stretch audio clips quickly with PyTorch (CUDA supported)! Additional utilities for searching efficient transformations are included.

Kento Nishi 22 Jul 07, 2022
Unsupervised CNN for Single View Depth Estimation: Geometry to the Rescue

Realtime Unsupervised Depth Estimation from an Image This is the caffe implementation of our paper "Unsupervised CNN for single view depth estimation:

Ravi Garg 227 Nov 28, 2022
ISTR: End-to-End Instance Segmentation with Transformers (https://arxiv.org/abs/2105.00637)

This is the project page for the paper: ISTR: End-to-End Instance Segmentation via Transformers, Jie Hu, Liujuan Cao, Yao Lu, ShengChuan Zhang, Yan Wa

Jie Hu 182 Dec 19, 2022
🍅🍅🍅YOLOv5-Lite: lighter, faster and easier to deploy. Evolved from yolov5 and the size of model is only 1.7M (int8) and 3.3M (fp16). It can reach 10+ FPS on the Raspberry Pi 4B when the input size is 320×320~

YOLOv5-Lite：lighter, faster and easier to deploy Perform a series of ablation experiments on yolov5 to make it lighter (smaller Flops, lower memory, a

pogg 1.5k Jan 05, 2023
Fast Axiomatic Attribution for Neural Networks (NeurIPS*2021)

Fast Axiomatic Attribution for Neural Networks This is the official repository accompanying the NeurIPS 2021 paper: R. Hesse, S. Schaub-Meyer, and S.

Visual Inference Lab @TU Darmstadt 11 Nov 21, 2022