Official Pytorch implementation for AAAI2021 paper (RSPNet: Relative Speed Perception for Unsupervised Video Representation Learning)

Related tags

Deep LearningRSPNet
Overview

RSPNet

Official Pytorch implementation for AAAI2021 paper "RSPNet: Relative Speed Perception for Unsupervised Video Representation Learning"

[Supplementary Materials]

Getting Started

Install Dependencies

All dependencies can be installed using pip:

python -m pip install -r requirements.txt

Our experiments run on Python 3.7 and PyTorch 1.6. Other versions should work but are not tested.

Transcode Videos (Optional)

This step is optional but will increase the data loading speed dramatically.

We decode the videos on the fly while training so we don't need to split frames. This makes disk IO a lot faster but increases CPU usage. This transcode step aims at reducing CPU consumed by decoding by 1) lower video resolution. 2) add more key frames.

To perform transcode, you need to have ffmpeg installed, then run:

python utils/transcode_dataset.py PATH/TO/ORIGIN_VIDEOS PATH/TO/TRANSCODED_VIDEOS

Be warned, this will use all your CPU and will take several hours (on our Intel E5-2630 *2 workstation) to complete.

Prepare Datasets

Your are expected to prepare date for pre-training (Kinetics-400 dataset) and fine-tuning (UCF101, HMDB51 and Something-something-v2 datasets). To let the scripts find datasets on your system, the recommended way is to create symbolic links in ./data directory to the actual path. We found this solution flexible.

The expected directory hierarchy is as follow:

├── data
│   ├── hmdb51
│   │   ├── metafile
│   │   │   ├── brush_hair_test_split1.txt
│   │   │   └── ...
│   │   └── videos
│   │       ├── brush_hair
│   │       │   └── *.avi
│   │       └── ...
│   ├── UCF101
│   │   ├── ucfTrainTestlist
│   │   │   ├── classInd.txt
│   │   │   ├── testlist01.txt
│   │   │   ├── trainlist01.txt
│   │   │   └── ...
│   │   └── UCF-101
│   │       ├── ApplyEyeMakeup
│   │       │   └── *.avi
│   │       └── ...
│   ├── kinetics400
│   │   ├── train_video
│   │   │   ├── answering_questions
│   │   │   │   └── *.mp4
│   │   │   └── ...
│   │   └── val_video
│   │       └── (same as train_video)
│   ├── kinetics100
│   │   └── (same as kinetics400)
│   └── smth-smth-v2
│       ├── 20bn-something-something-v2
│       │   └── *.mp4
│       └── annotations
│           ├── something-something-v2-labels.json
│           ├── something-something-v2-test.json
│           ├── something-something-v2-train.json
│           └── something-something-v2-validation.json
└── ...

Alternatively, you can change the path in config/dataset to match your system.

Build Kinetics-100 dataset (Optional)

Some of our ablation study experiments use the Kinetics-100 dataset for pre-training. This dataset is built by extract 100 classes from Kinetics-400, which has the smallest file size on the train set.

If you have Kinetics-400 available, you can build Kinetics-100 by:

python -m utils.build_kinetics_subset

This script will create symbolic links instead of copy data. It is expected to complete in a minute.

We have included a pre-built one at data/kinetics100_links and created the symbolic link data/kinetics100 that related to it. You need to have data/kinetics400 available at runtime.

Pre-training on Pretext Tasks

Now you have set up the environment. Run the following command to pre-train your models on pretext tasks.

export CUDA_VISIBLE_DEVICES=0,1,2,3
# Architecture: C3D
python pretrain.py -e exps/pretext-c3d -c config/pretrain/c3d.jsonnet
# Architecture: ResNet-18
python pretrain.py -e exps/pretext-resnet18 -c config/pretrain/resnet18.jsonnet
# Architecture: S3D-G
python pretrain.py -e exps/pretext-s3dg -c config/pretrain/s3dg.jsonnet
# Architecture: R(2+1)D
python pretrain.py -e exps/pretext-r2plus1d -c config/pretrain/r2plus1d.jsonnet

You can use kinetics100 dataset for training by editing config/pretrain/moco-train-base.jsonnet (line 13)

Action Recognition

After pre-trained on pretext tasks, these models are fine-tuned to perform action recognition task on UCF101, HMDB51 and Something-something-v2 datasets.

export CUDA_VISIBLE_DEVICES=0,1
# Dataset: UCF101
#     Architecture: C3D [email protected]=76.71%
python finetune.py -c config/finetune/ucf101_c3d.jsonnet \
                   --mc exps/pretext-c3d/model_best.pth.tar \
                   -e exps/ucf101-c3d
#     Architecture: ResNet-18 [email protected]=74.33%
python finetune.py -c config/finetune/ucf101_resnet18.jsonnet \
                   --mc exps/pretext-resnet18/model_best.pth.tar \
                   -e exps/ucf101-resnet18
#     Architecture: S3D-G [email protected]=89.9%
python finetune.py -c config/finetune/ucf101_s3dg.jsonnet \
                   --mc exps/pretext-s3dg/model_best.pth.tar \
                   -e exps/ucf101-s3dg
#     Architecture: R(2+1)D [email protected]=81.1%
python finetune.py -c config/finetune/ucf101_r2plus1d.jsonnet \
                   --mc exps/pretext-r2plus1d/model_best.pth.tar \
                   -e exps/ucf101-r2plus1d

# Dataset: HMDB51
#     Architecture: C3D [email protected]=44.58%
python finetune.py -c config/finetune/hmdb51_c3d.jsonnet \
                   --mc exps/pretext-c3d/model_best.pth.tar \
                   -e exps/hmdb51-c3d
#     Architecture: ResNet-18 [email protected]=41.83%
python finetune.py -c config/finetune/hmdb51_resnet18.jsonnet \
                   --mc exps/pretext-resnet18/model_best.pth.tar \
                   -e exps/hmdb51-resnet18
#     Architecture: S3D-G [email protected]=59.6%
python finetune.py -c config/finetune/hmdb51_s3dg.jsonnet \
                   --mc exps/pretext-s3dg/model_best.pth.tar \
                   -e exps/hmdb51-s3dg
#     Architecture: R(2+1)D [email protected]=44.6%
python finetune.py -c config/finetune/hmdb51_r2plus1d.jsonnet \
                   --mc exps/pretext-r2plus1d/model_best.pth.tar \
                   -e exps/hmdb51-r2plus1d

# Dataset: Something-something-v2
#     Architecture: C3D [email protected]=47.76%
python finetune.py -c config/finetune/smth_smth_c3d.jsonnet \
                   --mc exps/pretext-c3d/model_best.pth.tar \
                   -e exps/smthv2-c3d
#     Architecture: ResNet-18 [email protected]=44.02%
python finetune.py -c config/finetune/smth_smth_resnet18.jsonnet \
                   --mc exps/pretext-resnet18/model_best.pth.tar \
                   -e exps/smthv2-resnet18
#     Architecture: S3D-G [email protected]=55.03%
python finetune.py -c config/finetune/smth_smth_s3dg.jsonnet \
                   --mc exps/pretext-s3dg/model_best.pth.tar \
                   -e exps/smthv2-s3dg

Results and Pre-trained Models

Architecture Pre-trained dataset Pre-training epoch Pre-trained model Acc. on UCF101 Acc. on HMDB51
S3D-G Kinetics-400 1000 Download link 93.7 64.7
S3D-G Kinetics-400 200 Download link 89.9 59.6
R(2+1)D Kinetics-400 200 Download link 81.1 44.6
ResNet-18 Kinetics-400 200 Download link 74.3 41.8
C3D Kinetics-400 200 Download link 76.7 44.6

Video Retrieval

The pretrained model can also be used in searching relevant videos based on the given query video.

export CUDA_VISIBLE_DEVICES=0 # use single GPU 
python retrieval.py -c config/retrieval/ucf101_resnet18.jsonnet \
                    --mc exps/pretext-resnet18/model_best.pth.tar \
                    -e exps/retrieval-resnet18    

The video retrieval result in our paper

Architecture k=1 k=5 k=10 k=20 k=50
C3D 36.0 56.7 66.5 76.3 87.7
ResNet-18 41.1 59.4 68.4 77.8 88.7

Visualization

We further visualize the region of interest (RoI) that contributes most to the similarity score using the class activation map (CAM) technique.

export CUDA_VISIBLE_DEVICES=0,1
python visualization.py -c config/pretrain/s3dg.jsonnet \
                        --load-model exps/pretext-s3dg/model_best.pth.tar \
                        -e exps/visual-s3dg \
                        -x '{batch_size: 1}'

The cam visualization results will be plotted in png files like

Troubleshoot

  • DECORDError cannot find video stream with wanted index: -1

    Some video from Kinetics dataset does not contain a valid video stream for some unknown reason. To filter them out, run python utils/verify_video.py PATH/TO/VIDEOS, then copy the output to the blacklist config in config/dataset/kinetics{400,100}.libsonnet. You need to have ffmpeg installed.

Citation

Please cite the following paper if you feel RSPNet useful to your research

@InProceedings{chen2020RSPNet,
author = {Peihao Chen, Deng Huang, Dongliang He, Xiang Long, Runhao Zeng, Shilei Wen, Mingkui Tan, and Chuang Gan},
title = {RSPNet: Relative Speed Perception for Unsupervised Video Representation Learning},
booktitle = {The AAAI Conference on Artificial Intelligence (AAAI)},
year = {2021}
}

Contact

For any question, please file an issue or contact

Peihao Chen: [email protected]
Deng Huang: [email protected]
Comments
  • r(2+1) d -18 pretrained model not fully reproducible

    r(2+1) d -18 pretrained model not fully reproducible

    Hi, I finetuned the given pre-trained r(2+1)d model on ucf-101 using the given finetuning code. It only achieves (76 -77%) accuracy. Can you confirm if the given model is the correct one. I use the same setup as mentioned in the readme.

    opened by fmthoker 3
  • framework image

    framework image

    hello, thank you for your great work. it's so smart idea!

    can you explain about framework image? i understand about RSP task, A-VID task is learned in 1 iteration. i think that it means 'anchor is same'. and i saw the algorithm, just sampling K clips in video V\v+, however, in paper fig 2. two clips in video, 1x clip and 2x clip 's features(green color) are going to g_a header and do contrastive learning. i think about you want to show us randomly selected speed.... is right? in real experiment, just c_i, c_j, {c_n}(K) clips in there? not 2K?

    thank you

    opened by youwantsy 2
  • The pre-training model of s3d-g model based on Imagenet and dynamics-400 data set?

    The pre-training model of s3d-g model based on Imagenet and dynamics-400 data set?

    Where can I download the pre training model of s3d-g model based on Imagenet and dynamics-400 dataset? Or can you upload it to this repository? 请问哪里可以下载到基于ImageNet和Kinetics-400数据集的S3D-G模型的预训练模型?或者请问作者可以上传一下公开吗?

    opened by LiangSiyv 2
  • Question about computational resources

    Question about computational resources

    Hi, Thanks for your wonderful paper and code. I want to know the computational resources of your experiments. 1. What and how many GPUs you use? 2. The training time of pretraining on K400 for 200 epochs. 3. The training time of finetuning on UCF101, HMDB51, Something-V2, respectively. Looking forward to your reply. Thanks.

    opened by wjn922 2
  • 'No configuration setting found for key force_n_crop'

    'No configuration setting found for key force_n_crop'

    I downloaded your S3D-G pre-trained model for my action recognition task on UCF101 but I keep getting this error:

    argument type: <class 'str'> Setting ulimit -n 8192 world_size=1 Using dist_url=tcp://127.0.0.1:36879 Local Rank: 0 2021-12-30 07:31:39,148|INFO |Args = Args(parser=None, config='config/finetune/ucf101_s3dg.jsonnet', ext_config=[], debug=False, experiment_dir=PosixPath('exps/ucf101-s3dg'), _run_dir=PosixPath('exps/ucf101-s3dg/run_2_20211230_073138'), load_checkpoint=None, load_model=None, validate=False, moco_checkpoint='exps/pretext-s3dg/model_best_s3dg_200epoch.pth.tar', seed=None, world_size=1, _continue=False, no_scale_lr=False) 2021-12-30 07:31:39,149|INFO |cudnn.benchmark = True 2021-12-30 07:31:39,278|INFO |Config = batch_size = 4 dataset { annotation_path = "data/UCF101/ucfTrainTestlist" fold = 1 mean = [ 0.485 0.456 0.406 ] name = "ucf101" num_classes = 101 root = "data/UCF101/UCF-101" std = [ 0.229 0.224 0.225 ] } final_validate { batch_size = 4 } log_interval = 10 method = "from-scratch" model { arch = "s3dg" } model_type = "multitask" num_epochs = 50 num_workers = 8 optimizer { dampening = 0 lr = 0.005 milestones = [ 50 100 150 ] momentum = 0.9 nesterov = false patience = 10 schedule = "cosine" weight_decay = 0.0001 } spatial_transforms { color_jitter { brightness = 0 contrast = 0 hue = 0 saturation = 0 } crop_area { max = 1 min = 0.25 } gray_scale = 0 size = 224 } temporal_transforms { frame_rate = 25 size = 64 strides = [ { stride = 1 weight = 1 } ] validate { final_n_crop = 10 n_crop = 1 stride = 1 } } validate { batch_size = 4 } 2021-12-30 07:31:39,282|INFO |Using global get_model_class({'arch': 's3dg'}) 2021-12-30 07:31:39,283|INFO |Using MultiTask Wrapper 2021-12-30 07:31:39,283|WARNING |<class 'moco.split_wrapper.MultiTaskWrapper'> using groups: 1 2021-12-30 07:31:39,383|INFO |Found fc: fc with in_features: 1024 2021-12-30 07:31:42,488|INFO |Building Dataset: VID: False, Split=train 2021-12-30 07:31:42,488|INFO |Temporal transform type: clip Traceback (most recent call last): File "finetune.py", line 502, in main() File "finetune.py", line 498, in main mp.spawn(main_worker, args=(args, dist_url,), nprocs=args.world_size) File "/home/ubuntu/anaconda3/envs/ucf101/lib/python3.8/site-packages/torch/multiprocessing/spawn.py", line 200, in spawn return start_processes(fn, args, nprocs, join, daemon, start_method='spawn') File "/home/ubuntu/anaconda3/envs/ucf101/lib/python3.8/site-packages/torch/multiprocessing/spawn.py", line 158, in start_processes while not context.join(): File "/home/ubuntu/anaconda3/envs/ucf101/lib/python3.8/site-packages/torch/multiprocessing/spawn.py", line 119, in join raise Exception(msg) Exception:

    -- Process 0 terminated with the following error: Traceback (most recent call last): File "/home/ubuntu/anaconda3/envs/ucf101/lib/python3.8/site-packages/torch/multiprocessing/spawn.py", line 20, in _wrap fn(i, *args) File "/home/ubuntu/RSPNet/finetune.py", line 452, in main_worker engine = Engine(args, cfg, local_rank=local_rank) File "/home/ubuntu/RSPNet/finetune.py", line 171, in init self.train_loader = self.data_loader_factory.build( File "/home/ubuntu/RSPNet/datasets/classification/init.py", line 81, in build temporal_transform = self.get_temporal_transform(split) File "/home/ubuntu/RSPNet/datasets/classification/init.py", line 276, in get_temporal_transform if tt_cfg.get_bool("force_n_crop"): File "/home/ubuntu/anaconda3/envs/ucf101/lib/python3.8/site-packages/pyhocon/config_tree.py", line 310, in get_bool string_value = self.get_string(key, default) File "/home/ubuntu/anaconda3/envs/ucf101/lib/python3.8/site-packages/pyhocon/config_tree.py", line 221, in get_string value = self.get(key, default) File "/home/ubuntu/anaconda3/envs/ucf101/lib/python3.8/site-packages/pyhocon/config_tree.py", line 209, in get return self._get(ConfigTree.parse_key(key), 0, default) File "/home/ubuntu/anaconda3/envs/ucf101/lib/python3.8/site-packages/pyhocon/config_tree.py", line 151, in _get raise ConfigMissingException(u"No configuration setting found for key {key}".format(key='.'.join(key_path[:key_index + 1]))) pyhocon.exceptions.ConfigMissingException: 'No configuration setting found for key force_n_crop'

    opened by aloma85 0
Releases(pretrained_model)
Unadversarial Examples: Designing Objects for Robust Vision

Unadversarial Examples: Designing Objects for Robust Vision This repository contains the code necessary to replicate the major results of our paper: U

Microsoft 93 Nov 28, 2022
simple artificial intelligence utilities

Simple AI Project home: http://github.com/simpleai-team/simpleai This lib implements many of the artificial intelligence algorithms described on the b

921 Dec 08, 2022
Source code of "Hold me tight! Influence of discriminative features on deep network boundaries"

Hold me tight! Influence of discriminative features on deep network boundaries This is the source code to reproduce the experiments of the NeurIPS 202

EPFL LTS4 19 Dec 10, 2021
Repository for code and dataset for our EMNLP 2021 paper - “So You Think You’re Funny?”: Rating the Humour Quotient in Standup Comedy.

AI-OpenMic Dataset The dataset is available for download via the follwing link. Repository for code and dataset for our EMNLP 2021 paper - “So You Thi

6 Oct 26, 2022
PyTorch implementation of ICLR 2022 paper PiCO: Contrastive Label Disambiguation for Partial Label Learning

PiCO: Contrastive Label Disambiguation for Partial Label Learning This is a PyTorch implementation of ICLR 2022 Oral paper PiCO; also see our Project

王皓波 147 Jan 07, 2023
Axel - 3D printed robotic hands and they controll with Raspberry Pi and Arduino combo

Axel It's our graduation project about 3D printed robotic hands and they control

0 Feb 14, 2022
Motion and Shape Capture from Sparse Markers

MoSh++ This repository contains the official chumpy implementation of mocap body solver used for AMASS: AMASS: Archive of Motion Capture as Surface Sh

Nima Ghorbani 135 Dec 23, 2022
TensorFlow implementation of the algorithm in the paper "Decoupled Low-light Image Enhancement"

Decoupled Low-light Image Enhancement Shijie Hao1,2*, Xu Han1,2, Yanrong Guo1,2 & Meng Wang1,2 1Key Laboratory of Knowledge Engineering with Big Data

17 Apr 25, 2022
Official Implementation of LARGE: Latent-Based Regression through GAN Semantics

LARGE: Latent-Based Regression through GAN Semantics [Project Website] [Google Colab] [Paper] LARGE: Latent-Based Regression through GAN Semantics Yot

83 Dec 06, 2022
A graphical Semi-automatic annotation tool based on labelImg and Yolov5

💕YOLOV5 semi-automatic annotation tool (Based on labelImg)

EricFang 247 Jan 05, 2023
ArcaneGAN by Alex Spirin

ArcaneGAN by Alex Spirin

Alex 617 Dec 28, 2022
Code of the paper "Multi-Task Meta-Learning Modification with Stochastic Approximation".

Multi-Task Meta-Learning Modification with Stochastic Approximation This repository contains the code for the paper "Multi-Task Meta-Learning Modifica

Andrew 3 Jan 05, 2022
A Jinja extension (compatible with Flask and other frameworks) to compile and/or compress your assets.

A Jinja extension (compatible with Flask and other frameworks) to compile and/or compress your assets.

Jayson Reis 94 Nov 21, 2022
NeurIPS 2021 paper 'Representation Learning on Spatial Networks' code

Representation Learning on Spatial Networks This repository is the official implementation of Representation Learning on Spatial Networks. Training Ex

13 Dec 29, 2022
Learning Chinese Character style with conditional GAN

zi2zi: Master Chinese Calligraphy with Conditional Adversarial Networks Introduction Learning eastern asian language typefaces with GAN. zi2zi(字到字, me

Yuchen Tian 2.2k Jan 02, 2023
This implementation contains the application of GPlearn's symbolic transformer on a commodity futures sector of the financial market.

GPlearn_finiance_stock_futures_extension This implementation contains the application of GPlearn's symbolic transformer on a commodity futures sector

Chengwei <a href=[email protected]"> 189 Dec 25, 2022
git《USD-Seg:Learning Universal Shape Dictionary for Realtime Instance Segmentation》(2020) GitHub: [fig2]

USD-Seg This project is an implement of paper USD-Seg:Learning Universal Shape Dictionary for Realtime Instance Segmentation, based on FCOS detector f

Ruolin Ye 80 Nov 28, 2022
PyTorch implementation of DirectCLR from paper Understanding Dimensional Collapse in Contrastive Self-supervised Learning

DirectCLR DirectCLR is a simple contrastive learning model for visual representation learning. It does not require a trainable projector as SimCLR. It

Meta Research 49 Dec 21, 2022
PyTorch implementation of "Contrast to Divide: self-supervised pre-training for learning with noisy labels"

Contrast to Divide: self-supervised pre-training for learning with noisy labels This is an official implementation of "Contrast to Divide: self-superv

55 Nov 23, 2022
Official PyTorch implementation of the paper "TEMOS: Generating diverse human motions from textual descriptions"

TEMOS: TExt to MOtionS Generating diverse human motions from textual descriptions Description Official PyTorch implementation of the paper "TEMOS: Gen

Mathis Petrovich 187 Dec 27, 2022