FV2ES: A Fully End2End Multimodal System for Fast Yet Effective Video Emotion Recognition, by Qinglan Wei, Xuling Huang, Yuan Zhang.
In the latest social networks, more and more people prefer to express their emotions in videos through text, speech, and rich facial expressions. Multimodal video emotion analysis techniques can help understand users’ inner world automatically based on human expressions and gestures in images, tones in voices, and recognized natural language. However, in the existing research, the acoustic modality has long been in a marginal position as compared to visual and textual modalities. That is, it tends to be more difficult to improve the contribution of the acoustic modality for the whole multimodal emotion recognition task. Besides, although better performance can be obtained by introducing common deep learning methods, the complex structures of these training models always result in low inference efficiency, especially when exposed to high-resolution and long-length videos. Moreover, the lack of a fully end-to-end multimodal video emotion recognition system hinders its application. In this paper, we designed a fully multimodal video-to-emotion system (named FV2ES) for fast yet effective recognition inference, whose benefits are threefold: (1) The adoption of the hierarchical attention method upon the sound spectra breaks through the limited contribution of the acoustic modality, and outperforms the existing models’ performance on both IEMOCAP and CMU-MOSEI datasets; (2) the introduction of the idea of multi-scale for visual extraction while single-branch for inference brings higher efficiency and maintains the prediction accuracy at the same time; (3) the further integration of data pre-processing into the aligned multimodal learning model allows the significant reduction of computational costs and storage space.
If you work is inspired by our paper or code, please cite it, thanks!
@inproceedings{dai-etal-2021-multimodal,
title = "FV2ES: A Fully End2End Multimodal System for Fast Yet Effective Video Emotion Recognition",
author = "Qinglan Wei and
Xuling Huang and
Yuan Zhang",
abstract = "In the latest social networks, more and more people prefer to express their emotions in videos through text, speech, and rich facial expressions. Multimodal video emotion analysis techniques can help understand users’ inner world automatically based on human expressions and gestures in images, tones in voices, and recognized natural language. However, in the existing research, the acoustic modality has long been in a marginal position as compared to visual and textual modalities. That is, it tends to be more difficult to improve the contribution of the acoustic modality for the whole multimodal emotion recognition task. Besides, although better performance can be obtained by introducing common deep learning methods, the complex structures of these training models always result in low inference efficiency, especially when exposed to high-resolution and long-length videos. Moreover, the lack of a fully end-to-end multimodal video emotion recognition system hinders its application. In this paper, we designed a fully multimodal video-to-emotion system (named FV2ES) for fast yet effective recognition inference, whose benefits are threefold: (1) The adoption of the hierarchical attention method upon the sound spectra breaks through the limited contribution of the acoustic modality, and outperforms the existing models’ performance on both IEMOCAP and CMU-MOSEI datasets; (2) the introduction of the idea of multi-scale for visual extraction while single-branch for inference brings higher efficiency and maintains the prediction accuracy at the same time; (3) the further integration of data pre-processing into the aligned multimodal learning model allows the significant reduction of computational costs and storage space.",
}
As mentioned in our paper, there are two public datasets used in our experiments, including the IEMOCAP and the CMU-MOSEI datasets.
The IEMOCAP dataset consists of multimodal data of three modalities of video, audio, and text transcription. We select six main categories from the original emotions: anger, happiness, excitement, sadness, frustration, and neutral. And to create a new split for the dataset, we randomly assign 70%, 10%, and 20% of the data to the training, validation, and test sets respectively.
The CMU-MOSEI dataset also consists of multimodal data of three modalities of vision, audio, and text. Six kinds of labels including happiness, sadness, anger, fear, disgust, and surprise were annotated for the videos. And the dataset contains 250 topics, 3837 videos, 23453 sentences, 1000 narrators, and the total duration reaches 65 hours.
The raw data can be downloaded from CMU-MOSEI (~120GB) and IEMOCAP (~16.5GB). However, for the IEMOCAP, you need to request for a permission from the original author, then you can be given the passcode to download.
To run our code directly, you can download the processed data from here (88.6G). Unzip it and the tree structure of the data direcotry looks like this:
./data
- IEMOCAP_HCF_FEATURES
- IEMOCAP_RAW_PROCESSED
- IEMOCAP_SPLIT
- MOSEI_RAW_PROCESSED
- MOSEI_HCF_FEATURES
- MOSEI_SPLIT
- Python 3.7.6
- PyTorch 1.8.0
- torchaudio 0.8.0
- torchvision 0.9.0
- transformers 4.17.0
- facenet-pytorch 2.5.2
python main.py -lr=4.5e-6 -ep=30 -mod=tav -bs=2 --img-interval=500 --early-stop=6 --loss=bce --cuda=0 --model=mme2e --num-emotions=6 --trans-dim=64 --trans-nlayers=4 --trans-nheads=4 --text-lr-factor=10 --text-model-size=base --text-max-len=100 You can start the system by running FV2ES/System/app.py
usage: main.py [-h] -bs BATCH_SIZE -lr LEARNING_RATE [-wd WEIGHT_DECAY] -ep
EPOCHS [-es EARLY_STOP] [-cu CUDA] [-cl CLIP] [-sc] [-se SEED]
[--loss LOSS] [--optim OPTIM] [--text-lr-factor TEXT_LR_FACTOR]
[-mo MODEL] [--text-model-size TEXT_MODEL_SIZE]
[--fusion FUSION] [--feature-dim FEATURE_DIM] [-hfcs HFC_SIZES [HFC_SIZES ...]]
[--trans-dim TRANS_DIM] [--trans-nlayers TRANS_NLAYERS]
[--trans-nheads TRANS_NHEADS] [-aft AUDIO_FEATURE_TYPE]
[--num-emotions NUM_EMOTIONS] [--img-interval IMG_INTERVAL]
[--hand-crafted] [--text-max-len TEXT_MAX_LEN]
[--datapath DATAPATH] [--dataset DATASET] [-mod MODALITIES]
[--valid] [--test] [--ckpt CKPT] [--ckpt-mod CKPT_MOD]
[-dr DROPOUT] [-nl NUM_LAYERS] [-hs HIDDEN_SIZE] [-bi] [--gru]
FV2ES: A Fully End2End Multimodal System for Fast Yet Effective Video Emotion Recognition
optional arguments:
-h, --help show this help message and exit
-bs BATCH_SIZE, --batch-size BATCH_SIZE
Batch size
-lr LEARNING_RATE, --learning-rate LEARNING_RATE
Learning rate
-wd WEIGHT_DECAY, --weight-decay WEIGHT_DECAY
Weight decay
-ep EPOCHS, --epochs EPOCHS
Number of epochs
-es EARLY_STOP, --early-stop EARLY_STOP
Early stop
-cu CUDA, --cuda CUDA
Cude device number
-cl CLIP, --clip CLIP
Use clip to gradients
-sc, --scheduler Use scheduler to optimizer
-se SEED, --seed SEED
Random seed
--loss LOSS loss function
--optim OPTIM optimizer function: adam/sgd
--text-lr-factor TEXT_LR_FACTOR
Factor the learning rate of text model
-mo MODEL, --model MODEL
Which model
--text-model-size TEXT_MODEL_SIZE
Size of the pre-trained text model
--fusion FUSION How to fuse modalities
--feature-dim FEATURE_DIM
Dimension of features outputed by each modality model
-hfcs HFC_SIZES [HFC_SIZES ...], --hfc-sizes HFC_SIZES [HFC_SIZES ...]
Hand crafted feature sizes
--trans-dim TRANS_DIM
Dimension of the transformer after CNN
--trans-nlayers TRANS_NLAYERS
Number of layers of the transformer after CNN
--trans-nheads TRANS_NHEADS
Number of heads of the transformer after CNN
-aft AUDIO_FEATURE_TYPE, --audio-feature-type AUDIO_FEATURE_TYPE
Hand crafted audio feature types
--num-emotions NUM_EMOTIONS
Number of emotions in data
--img-interval IMG_INTERVAL
Interval to sample image frames
--hand-crafted Use hand crafted features
--text-max-len TEXT_MAX_LEN
Max length of text after tokenization
--datapath DATAPATH Path of data
--dataset DATASET Use which dataset
-mod MODALITIES, --modalities MODALITIES
what modalities to use
--valid Only run validation
--test Only run test
--ckpt CKPT Path of checkpoint
--ckpt-mod CKPT_MOD Load which modality of the checkpoint
-dr DROPOUT, --dropout DROPOUT
dropout
-nl NUM_LAYERS, --num-layers NUM_LAYERS
num of layers of LSTM
-hs HIDDEN_SIZE, --hidden-size HIDDEN_SIZE
hidden vector size of LSTM
-bi, --bidirectional Use Bi-LSTM
--gru Use GRU rather than LSTM
please download from https://pan.baidu.com/s/1lpjAMjLrPy-HNHAZIrCR1g?pwd=2930 code:2930