Awesome
EgoVLP: Egocentric Video-Language Pretraining
TL;DR: We pioneer Egocentric Video-Language Pretraining from pretraining dataset, model and development benchmark; the resulted pretrained model exhibits strong performance on five downstream tasks across three egocentric datasets.
📢 News
- [2023.7.12] EgoVLPv2 has been released, directed by Shraman, with stronger performance and higher efficiency, which has been accepted by ICCV 2023!
- [2022.12.22] We clean the code and provide video features to power NLQ & MQ, Ego4D challenges.
- [2022.9.15] EgoVLP got accepted by NeurIPS 2022 as Spotlight!
- [2022.6.30] We release the first version of the EgoVLP codebase.
- [2022.6.20] Our EgoVLP won 1st place in OSCC & 2nd place in NLQ & 3rd place in PNR @ Ego4D Challenge 2022, and 1st place in Multi-Instance Retrieval @ EPIC-Kitchens Challenge 2022, hosted by CVPR 2022.
- [2022.6.10] We release the EgoClip pretraining dataset.
- [2022.6.3] We release the arXiv paper.
📝 Preparation
Install dependencies
conda env create -f environment.yml
source activate egovlp
Ego4D videos and metadata
You can skip the source video download if pretraining is not required.
-
Follow the guideline here, download the following to
{PATH_TO_EGO4D}
- Ego4D source videos (nearly 7 TB).
- Ego4D videos metadata
manifest.csv
and benchmark metadata, e.g.,nlq_train.json
for NLQ. - Create the dir
dataset
and add a soft link byln -s {PATH_TO_EGO4D} dataset/ego4d
.
-
For effectively pretraining, we compress videos in the following way:
- Resize the source videos with a short size equal to 256 by script
utils/video_resize.py
. - Chunk the resized videos to multiple segments (up to 600 sec) by script
utils/video_chunk.py
.
- Resize the source videos with a short size equal to 256 by script
EgoClip: an egocentric video-language pretraining dataset
-
Download the EgoClip metadata from here and put it to
dataset/egoclip.csv
. -
For the usage of EgoClip, please see our dataloader
data_loader/EgoClip_EgoMCQ_dataset.py
. The data format of EgoClip is:import pandas as pd metadata = pd.read_csv('dataset/egoclip.csv', sep='\t', error_bad_lines=False) print(metadata.shape[0]) print(metadata.iloc[0]) # Out: 3847723 # Num of clips for EgoClip clip_idx 0 # the idx of clip video_uid 001e3e4e-2743-47fc-8564-d5efd11f9e90 # the uid of source video video_dur 128.033333 # the duration of source video narration_source narration_pass_1 # the source of annotator narration_ind 0 # the idx of narration narration_time 3.3445 # the narration timestamp clip_start 2.967651 # the start timestamp of clip clip_end 3.721266 # the end timestamp of clip clip_text #C C picks a bag of clothes from the floor # the narration of clip tag_verb [93] # the verb idx of the narration tag_noun [192, 115, 12] # the noun idx of the narration
^ The terms tag_verb
and tag_noun
are used for EgoNCE pretraining objective, which considers synonyms. For example, pick
, collect
, gather
are all belong to the verb parent with idx 93: take_(pick,_grab,_get)
.
The mapping dictionary can be found here.
EgoMCQ: an egocentric video-language development set
- Download the EgoMCQ metadata from here and put it to
dataset/egomcq.json
. - EgoMCQ is a benchmark for video-language multiple-choice questions. Given a text query, we want the model to choose the correct video clip from five candidates that sampled from two settings:
inter-video
orintra-video
. - For the usage of EgoMCQ, please see our dataloader
data_loader/EgoClip_EgoMCQ_dataset.py
.
🏋️️ Pretraining
This code is built on PyTorch with DistributedDataParallel (DDP). We pretrain EgoVLP on 4 nodes, each with 8 A100 GPUs (10 epochs in about two days).
-
Train on EgoClip:
python3 -m torch.distributed.launch --nnodes=$HOST_NUM --node_rank=$INDEX --master_addr $CHIEF_IP --nproc_per_node $HOST_GPU_NUM --master_port 8081 run/train_egoclip.py --config configs/pt/egoclip.json
-
Test on EgoMCQ:
python3 -m torch.distributed.launch --nnodes=$HOST_NUM --node_rank=$INDEX --master_addr $CHIEF_IP --nproc_per_node $HOST_GPU_NUM --master_port 8081 run/train_egoclip.py --config configs/eval/egomcq.json
-
Monitor the EgoMCQ curve during pretraining:
tensorboard --logdir results --bind_all
🗄 Pretrained Weights
- We have released our pretrained EgoVLP model (EgoClip w/ EgoNCE) with best performance on EgoMCQ (90.7% inter-video & 57.2% intra-video) in EgoVLP_PT_BEST.
- Please download and put the checkpoint under:
pretrained/
^ This checkpoint is used for EPIC-Kitchens, NLQ, MQ, OSSC, and PNR tasks, except for Charades-Ego. Since we found that VLP (CC3M+WebVid2M, EgoClip) alway degrades significantly on Charades-Ego after the first epoch, we evaluate Charades-Ego using the first pretraining epoch weights of EgoVLP in EgoVLP_PT_EPO1.
^^ You can use our checkpoint to power other egocentric video benchmarks. :)
🔧 Downstream Tasks
EPIC-Kitchens MIR
- Preparation:
- Follow the instruction here, download the EPIC-Kitchens dataset (RGB frames) and annotation to path:
dataset/epic-kitchens/
- Follow the instruction here -> How do I create the relevance matrix? to construct a relevance matrix for evaluation.
- Results:
Model | Mode | # Frames | Video-Text PT | Weights | mAP (V2T) | mAP (T2V) | mAP (Avg) | nDCG (V2T) | nDCG (T2V) | nDCG (Avg) |
---|---|---|---|---|---|---|---|---|---|---|
EgoVLP | Zero-shot | 4 | EgoClip w/ EgoNCE | EgoVLP_PT_BEST | 19.4 | 13.9 | 16.6 | 24.1 | 22.0 | 23.1 |
EgoVLP | Fine-tuning w/<br /> MI-MM | 16 | EgoClip w/ EgoNCE | EgoVLP_FT_EPIC | 49.9 | 40.5 | 45.0 | 60.9 | 57.9 | 59.4 |
EgoVLP+ | Fine-tuning w/ Adaptive-MI-MM + Dual-softmax | 16 | EgoClip w/ EgoNCE | EgoVLP_FT_EPIC+ | 53.8 | 40.9 | 47.4 | 63.3 | 59.6 | 61.4 |
^ EgoVLP+ means our submission for Multi-Instance Retrieval@EPIC-Kitchens Challenge 2022, which equips Adaptive MI-MM loss and Dual-softmax for prediction.
-
Train:
python3 -m torch.distributed.launch --nnodes=$HOST_NUM --node_rank=$INDEX --nproc_per_node $HOST_GPU_NUM --master_port 8081 run/train_epic.py --config configs/ft/epic.json
-
Test:
python3 run/test_epic.py
Charades-Ego
- Preparation:
- Follow the instruction here, download the Charades-Ego dataset (480p) and annotation to path:
dataset/charades/
- Create a training metadata via
utils/charades_meta.py
- Results:
Model | Mode | # Frames | Video-Text PT | Weights | mAP |
---|---|---|---|---|---|
EgoVLP | Zero-shot | 16 | EgoClip w/ EgoNCE | EgoVLP_PT_EPO1 | 25.0 |
EgoVLP | Fine-tuning w/ InfoNCE | 16 | EgoClip w/ EgoNCE | EgoVLP_FT_CHARADES | 32.1 |
-
Train:
python3 -m torch.distributed.launch --nnodes=$HOST_NUM --node_rank=$INDEX --nproc_per_node $HOST_GPU_NUM --master_port 8081 run/train_charades.py --config configs/ft/charades.json
-
Test:
python3 run/test_charades.py
NLQ @ Ego4D
- Preparation:
- Make sure you have prepared the NLQ metadata.
- For the video branch, download the EgoVLP clip-level features for NLQ. ^ We get these dense video features (fps=1.87) by script
run/test_nlq.py
. - For the text branch, you can extract EgoVLP text features:
python3 run/test_nlq.py --subsample 'text'
or use our pretrained text encoder. - Fine-tune the VSLNet or other methods by replacing their input video-text features.
^ We provide our VSLNet codebase which adapts EgoVLP features as an example, you can refer to the data loader and text encoder.
^ Our EgoVLP brings consistent improvement over multiple NLQ challenge baselines.
Model | Video-Text Pre-extrated Features | R@1, IoU=0.3 | R@5, IoU=0.3 | R@1, IoU=0.5 | R@5, IoU=0.5 |
---|---|---|---|---|---|
VSLNet | SlowFast + BERT | 5.45 | 10.74 | 3.12 | 6.63 |
VSLNet | EgoVLP | 10.84 | 18.84 | 6.81 | 13.45 |
CONE | SlowFast + BERT | 10.40 | 22.74 | 5.03 | 11.87 |
CONE | EgoVLP | 14.15 | 30.33 | 8.18 | 18.02 |
MQ @ Ego4D
- Preparation:
- Make sure you have prepared the MQ metadata.
- Download the EgoVLP clip-level features for MQ. ^ We get these dense video features (fps=1.87) by script
run/test_mq.py
. - Fine-tune the VSGN or other methods by replacing their input video features.
^ We provide our VSGN codebase which adapts EgoVLP features as an example, you can refer to the data loader.
^ Our EgoVLP brings consistent improvement over multiple MQ challenge baselines.
Model | Video Pre-extrated Features | R@1, IoU=0.5 | R@5, IoU=0.5 | mAP |
---|---|---|---|---|
VSGN | SlowFast | 25.16 | 46.18 | 6.03 |
VSGN | EgoVLP | 30.14 | 51.98 | 11.39 |
ActionFormer | SlowFast + Omnivore | 33.46 | - | 17.17 |
ActionFormer | SlowFast + Omnivore + EgoVLP | 36.84 | - | 20.90 |
OSCC @ Ego4D
- Preparation:
- Make sure you have prepared the OSCC videos and metadata.
- Extract the clip frame follow the instruction here -> Data Preparation.
- Train:
python3 -m torch.distributed.launch --nnodes=$HOST_NUM --node_rank=$INDEX --nproc_per_node $HOST_GPU_NUM --master_port 8081 run/train_oscc.py --config configs/ft/oscc.json
Model | Video-Text Pretrained | OSCC Acc % |
---|---|---|
TimeSformer | ImageNet Init. | 70.3 |
TimeSformer | EgoVLP | 73.9 |
PNR @ Ego4D
- Preparation: Same as OSCC.
- Train:
python3 -m torch.distributed.launch --nnodes=$HOST_NUM --node_rank=$INDEX --nproc_per_node $HOST_GPU_NUM --master_port 8081 run/train_pnr.py --config configs/ft/pnr.json
Model | Video-Text Pretrained | PNR Err % |
---|---|---|
TimeSformer | ImageNet Init. | 0.616 |
TimeSformer | EgoVLP | 0.622 |
^ We found VLP effect is minor in the PNR task.
🎓 Citation
If you find our work helps, please cite our paper.
@article{kevin2022egovlp,
title={Egocentric Video-Language Pretraining},
author={Lin, Kevin Qinghong and Wang, Alex Jinpeng and Soldan, Mattia and Wray, Michael and Yan, Rui and Xu, Eric Zhongcong and Gao, Difei and Tu, Rongcheng and Zhao, Wenzhe and Kong, Weijie and others},
journal={arXiv preprint arXiv:2206.01670},
year={2022}
}
✉️ Contact
This repo is maintained by Kevin. Questions and discussions are welcome via kevin.qh.lin@gmail.com
.
We are willing to merge results and codes if transfer our EgoVLP to other egocentric tasks or datasets.
🙏 Acknowledgements
This codebase is based on Frozen.
Thanks to Alex for the help with DDP and Mattia for the help with NLQ and MQ benchmarks.
LICENSE
MIT