Deep Learning Face Attributes in the Wild
By Ziwei Liu, Ping Luo, Xiaogang Wang, Xiaoou Tang. 2014
9,800 citations recorded by the source corpus.
Predicting face attributes in the wild is challenging due to complex face variations. We propose a novel deep learning framework for attribute prediction in the wild. It cascades two CNNs, LNet and ANet, which are fine-tuned jointly with attribute tags, but pre-trained differently. LNet is pre-trained by massive general object categories for face localization, while ANet is pre-trained by massive face identities for attribute prediction. This framework not only outperforms the state-of-the-art with a large margin, but also reveals valuable facts on learning face representation. (1) It shows how the performances of face localization (LNet) and attribute prediction (ANet) can be improved by different pre-training strategies. (2) It reveals that although the filters of LNet are fine-tuned only with image-level attribute tags, their response maps over entire images have strong indication of face locations. This fact enables training LNet for face localization with only image-level annotations, but without face bounding boxes or landmarks, which are required by all attribute recognition works. (3) It also demonstrates that the high-level hidden neurons of ANet automatically discover semantic concepts after pre-training with massive face identities, and such concepts are significantly enriched after fine-tuning with attribute tags. Each attribute can be well explained with a sparse linear combination of these concepts.
Subfields: Computer Vision and Pattern Recognition, Signal Processing
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Computer Vision and Pattern Recognition
- Luc Van Gool
- Dacheng Tao
- Ming-Hsuan Yang
- Trevor Darrell
- Xiaogang Wang
- Jitendra Malik
- Andrew Zisserman
- Andrea Vedaldi
In Signal Processing
- Yoshua Bengio
- Stefano Ermon
- Lior Wolf
- Andrew Zisserman
- Ramalingam Chellappa
- Aaron C. Courville
- Xiaoou Tang
- Dong Yu
Citations in the corpus
- InfoGAN: Interpretable Representation Learning by Information Maximizing Generative Adversarial Nets (2016, background): “Despite its simplicity, we found our method to be surprisingly effective: it was able to discover highly semantic and meaningful hidden representations on a number of image datasets: digits (MNIST), faces (CelebA), and house numbers (SVHN).”
- Experience Grounds Language (2020, background): “For example, facial recognition benchmarks require distinguishing over 10,000 unique faces (Liu et al., 2015).”
- Model Cards for Model Reporting (2018, methodology): “ms [5, 26] that report details of the datasets used to train and test machine learning models. We provide two example model cards in Section 5: A smiling detection model trained on the CelebA dataset [33] (Figure 2), and a public toxicity detection model, Perspective API [29] (Figure 3). Where Datas”
- Mixed Precision Training (2017, methodology): “As an example, consider Figure 3 showing the histogram of activation gradient values, collected across all layers during FP32 training of Multibox SSD detector network (Liu et al., 2015a).”
- On The Impact of Machine Learning Randomness on Group Fairness (2023, methodology): “[33] Ziwei Liu, Ping Luo, XiaogangWang, and Xiaoou Tang.”
- FairFace: Face Attribute Dataset for Balanced Race, Gender, and Age (2019, methodology): “To date, numerous large scale face image datasets [22, 31, 13, 69, 37, 24, 41, 68, 15, 27, 46, 8, 39] have been proposed and fostered research and development for automated face detection [35, 21], alignment [66, 44], recognition [57, 49], generation [67, 5, 26, 58], modification [3, 32, 19], and…”
- Improved Techniques for Training Score-Based Generative Models (2020, methodology): “As discussed before in Section 6, we note that FID scores should be interpreted with caution as they may not correlate with visual quality in the expected manner, so we also report HYPE ∞ scores, a metric based on human evaluation, on CelebA 64 × 64 .”
- VQ-DRAW: A Sequential Discrete VAE (2020, methodology): “I trained VQ-DRAW models on four different image datasets: MNIST [12], SVHN [13], CIFAR-10 [14], and CelebA [15].”
- Energy-Inspired Models: Learning with Sampler-Induced Distributions (2019, methodology): “We evaluated the proposed models on a set of synthetic datasets, binarized MNIST [42] and Fashion MNIST [67], and continuous MINST, Fashion MNIST, and CelebA [44].”
- Generative Modeling by Estimating Gradients of the Data Distribution (2019, methodology): “Experimentally, we demonstrate the efficacy of our approach on MNIST, CelebA [30], and CIFAR-10 [27].”
- Autoregressive Quantile Networks for Generative Modeling (2018, methodology): “We demonstrate the AIQN-VAE using the CelebA dataset (Liu et al., 2015), at resolution 64 × 64 . to We modified an open source VAE implementation 4 simultaneously train the AIQN on the output of the VAE encoder, with Polyak averaging (Polyak & Juditsky, 1992) of the AIQN weights.”
- Variational Walkback: Learning a Transition Operator as a Stochastic Recurrent Net (2017, methodology): “We then describe experimental results illustrating the soundness of the proposed approach on the MNIST, CIFAR-10, SVHN and CelebA datasets.”
- Progressive Growing of GANs for Improved Quality, Stability, and Variation (2017, background): “…methods that produce novel samples from high-dimensional data distributions, such as images, are finding widespread use, for example in speech synthesis (van den Oord et al., 2016a), image-to-image translation (Zhu et al., 2017; Liu et al., 2017), and image inpainting (Iizuka et al., 2017).”
- Learning to Generate Samples from Noise through Infusion Training (2017, methodology): “CelebA was trained with the previous network where we added another transposed convolution.”
- Density estimation using Real NVP (2016, methodology): “We train our model on four natural image datasets: CIFAR-10 [29], Imagenet [45], Large-scale Scene Understanding (LSUN) [62], CelebFaces Attributes (CelebA) [34].”
- On Fast Sampling of Diffusion Probabilistic Models (2021, methodology): “, 2009)), CelebA (∼163k face images of resolution 64× 64 (Liu et al., 2015)), and LSUN-bedroom (∼3M bedroom images of resolution 256× 256 (Yu et al.”
- Diffusion Models Beat GANs on Image Synthesis (2021, neutral): “Song et al. [60] found that further changes to the UNet architecture improved performance on the CIFAR-10 [31] and CelebA-64 [34] datasets.”
- Noise Estimation for Generative Diffusion Models (2021, methodology): “We trained our model Pθ on three image datasets (i) CelebA 64x64 [20], (ii) LSUN Bedroom 256x256, and (iii) LSUN Church 256x256 [36].”