no code implementations • 30 Jan 2024 • Mehdi Noroozi, Isma Hadji, Brais Martinez, Adrian Bulat, Georgios Tzimiropoulos
We show that the combination of spatially distilled U-Net and fine-tuned decoder outperforms state-of-the-art methods requiring 200 steps with only one single step.
no code implementations • 28 Jul 2023 • Ioannis Maniadis Metaxas, Adrian Bulat, Ioannis Patras, Brais Martinez, Georgios Tzimiropoulos
DETR-based object detectors have achieved remarkable performance but are sample-inefficient and exhibit slow convergence.
1 code implementation • ICCV 2023 • Yassine Ouali, Adrian Bulat, Brais Martinez, Georgios Tzimiropoulos
Vision-Language (V-L) models trained with contrastive learning to align the visual and language modalities have been shown to be strong few-shot learners.
no code implementations • ICCV 2023 • Adrian Bulat, Enrique Sanchez, Brais Martinez, Georgios Tzimiropoulos
Specifically, we propose ReGen, a novel reinforcement learning based framework with a three-fold objective and reward functions: (1) a class-level discrimination reward that enforces the generated caption to be correctly classified into the corresponding action class, (2) a CLIP reward that encourages the generated caption to continue to be descriptive of the input video (i. e. video-specific), and (3) a grammar reward that preserves the grammatical correctness of the caption.
no code implementations • ICCV 2023 • Adrian Bulat, Ricardo Guerrero, Brais Martinez, Georgios Tzimiropoulos
Importantly, we show that our system is not only more flexible than existing methods, but also, it makes a step towards satisfying desideratum (c).
1 code implementation • ICCV 2023 • Mohammad Mahdi Derakhshani, Enrique Sanchez, Adrian Bulat, Victor Guilherme Turrisi da Costa, Cees G. M. Snoek, Georgios Tzimiropoulos, Brais Martinez
Our approach regularizes the prompt space, reduces overfitting to the seen prompts and improves the prompt generalization on unseen prompts.
Ranked #1 on Few-Shot Learning on food101
1 code implementation • CVPR 2023 • Adrian Bulat, Georgios Tzimiropoulos
Through evaluations on 11 datasets, we show that our approach (a) significantly outperforms all prior works on soft prompting, and (b) matches and surpasses, for the first time, the accuracy on novel classes obtained by hand-crafted prompts and CLIP for 8 out of 11 test datasets.
no code implementations • 29 Sep 2022 • Adrian Bulat, Enrique Sanchez, Brais Martinez, Georgios Tzimiropoulos
We evaluate REST on the problem of zero-shot action recognition where we show that our approach is very competitive when compared to contrastive learning-based methods.
no code implementations • 23 Aug 2022 • Adrian Bulat, Brais Martinez, Georgios Tzimiropoulos
To address this gap, in this paper, we make the following contributions: (a) we construct a highly efficient \& accurate attention-free block based on the shift operator, coined Affine-Shift block, specifically designed to approximate as closely as possible the operations in the MHSA block of a Transformer layer.
no code implementations • 16 Jun 2022 • Fatemeh Saleh, Fuwen Tan, Adrian Bulat, Georgios Tzimiropoulos, Brais Martinez
Video self-supervised learning (SSL) suffers from added challenges: video datasets are typically not as large as image datasets, compute is an order of magnitude larger, and the amount of spurious patterns the optimizer has to sieve through is multiplied several fold.
1 code implementation • 13 May 2022 • Jing Yang, Xiatian Zhu, Adrian Bulat, Brais Martinez, Georgios Tzimiropoulos
The key idea is that we leverage the teacher's classifier as a semantic critic for evaluating the representations of both teacher and student and distilling the semantic knowledge with high-order structured information over all feature dimensions.
1 code implementation • 6 May 2022 • Junting Pan, Adrian Bulat, Fuwen Tan, Xiatian Zhu, Lukasz Dudziak, Hongsheng Li, Georgios Tzimiropoulos, Brais Martinez
In this work, pushing further along this under-studied direction we introduce EdgeViTs, a new family of light-weight ViTs that, for the first time, enable attention-based vision models to compete with the best light-weight CNNs in the tradeoff between accuracy and on-device efficiency.
no code implementations • 3 Nov 2021 • Adrian Bulat, Enrique Sanchez, Georgios Tzimiropoulos
Deep Learning models based on heatmap regression have revolutionized the task of facial landmark localization with existing models working robustly under large poses, non-uniform illumination and shadows, occlusions and self-occlusions, low resolution and blur.
Ranked #1 on Face Alignment on WFW (Extra Data) (using extra training data)
no code implementations • 26 Oct 2021 • Adrian Bulat, Jean Kossaifi, Sourav Bhattacharya, Yannis Panagakis, Timothy Hospedales, Georgios Tzimiropoulos, Nicholas D Lane, Maja Pantic
We propose defensive tensorization, an adversarial defence technique that leverages a latent high-order factorization of the network.
no code implementations • 6 Oct 2021 • Swathikiran Sudhakaran, Adrian Bulat, Juan-Manuel Perez-Rua, Alex Falcon, Sergio Escalera, Oswald Lanz, Brais Martinez, Georgios Tzimiropoulos
This report presents the technical details of our submission to the EPIC-Kitchens-100 Action Recognition Challenge 2021.
1 code implementation • NeurIPS 2021 • Adrian Bulat, Juan-Manuel Perez-Rua, Swathikiran Sudhakaran, Brais Martinez, Georgios Tzimiropoulos
In this work, we propose a Video Transformer model the complexity of which scales linearly with the number of frames in the video sequence and hence induces no overhead compared to an image-based Transformer model.
Ranked #32 on Action Classification on Kinetics-600
no code implementations • ICCV 2021 • Adrian Bulat, Georgios Tzimiropoulos
In this work, we propose Bit-Mixer, the very first method to train a meta-quantized network where during test time any layer can change its bid-width without affecting at all the overall network's ability for highly accurate inference.
2 code implementations • 30 Mar 2021 • Adrian Bulat, Shiyang Cheng, Jing Yang, Andrew Garbett, Enrique Sanchez, Georgios Tzimiropoulos
Recent work on Deep Learning in the area of face analysis has focused on supervised learning for specific tasks of interest (e. g. face recognition, facial landmark localization etc.)
Ranked #1 on Facial Expression Recognition (FER) on BP4D
no code implementations • 8 Feb 2021 • Adrian Bulat, Enrique Sánchez-Lozano, Georgios Tzimiropoulos
An important component of unsupervised learning by instance-based discrimination is a memory bank for storing a feature representation for each training sample in the dataset.
no code implementations • ICLR 2021 • Jing Yang, Brais Martinez, Adrian Bulat, Georgios Tzimiropoulos
We advocate for a method that optimizes the output feature of the penultimate layer of the student network and hence is directly related to representation learning.
no code implementations • 3 Nov 2020 • Enrique Sanchez, Adrian Bulat, Anestis Zaganidis, Georgios Tzimiropoulos
The second stage uses another dataset of randomly chosen labeled frames to train a regressor on top of our spatio-temporal model for estimating the AU intensity.
1 code implementation • ICLR 2021 • Adrian Bulat, Brais Martinez, Georgios Tzimiropoulos
Network binarization is a promising hardware-aware direction for creating efficient deep models.
no code implementations • 14 Apr 2020 • Ioanna Ntinou, Enrique Sanchez, Adrian Bulat, Michel Valstar, Georgios Tzimiropoulos
Action Units (AUs) are geometrically-based atomic facial muscle movements known to produce appearance changes at specific facial locations.
1 code implementation • ICLR 2020 • Brais Martinez, Jing Yang, Adrian Bulat, Georgios Tzimiropoulos
This paper shows how to train binary networks to within a few percent points ($\sim 3-5 \%$) of the full precision counterpart.
no code implementations • 9 Mar 2020 • Jing Yang, Brais Martinez, Adrian Bulat, Georgios Tzimiropoulos
To this end, we propose a new knowledge distillation method based on transferring feature statistics, specifically the channel-wise mean and variance, from the teacher to the student.
no code implementations • ECCV 2020 • Adrian Bulat, Brais Martinez, Georgios Tzimiropoulos
We show that directly applying NAS to the binary domain provides very poor results.
Ranked #1 on Classification with Binary Neural Network on CIFAR-10 (using extra training data)
3 code implementations • 25 Feb 2020 • Adrian Bulat, Jean Kossaifi, Georgios Tzimiropoulos, Maja Pantic
In addition, with a reduction of 3x in model size and complexity, we show no decrease in performance when compared to the original HourGlass network.
Ranked #2 on Pose Estimation on Leeds Sports Poses
no code implementations • 14 Nov 2019 • Shiyang Cheng, Pingchuan Ma, Georgios Tzimiropoulos, Stavros Petridis, Adrian Bulat, Jie Shen, Maja Pantic
The proposed model significantly outperforms previous approaches on non-frontal views while retaining the superior performance on frontal and near frontal mouth views.
1 code implementation • 30 Sep 2019 • Adrian Bulat, Georgios Tzimiropoulos
This paper proposes an improved training algorithm for binary neural networks in which both weights and activations are binary numbers.
no code implementations • 25 Sep 2019 • Adrian Bulat, Jean Kossaifi, Sourav Bhattacharya, Yannis Panagakis, Georgios Tzimiropoulos, Nicholas D. Lane, Maja Pantic
As deep neural networks become widely adopted for solving most problems in computer vision and audio-understanding, there are rising concerns about their potential vulnerability.
no code implementations • CVPR 2020 • Jean Kossaifi, Antoine Toisoul, Adrian Bulat, Yannis Panagakis, Timothy Hospedales, Maja Pantic
To alleviate this, one approach is to apply low-rank tensor decompositions to convolution kernels in order to compress the network and reduce its number of parameters.
no code implementations • 16 Apr 2019 • Adrian Bulat, Jean Kossaifi, Georgios Tzimiropoulos, Maja Pantic
This paper is on improving the training of binary neural networks in which both activations and weights are binary.
no code implementations • 12 Apr 2019 • Adrian Bulat, Jean Kossaifi, Georgios Tzimiropoulos, Maja Pantic
Adapting the learned classification to new domains is a hard problem due to at least three reasons: (1) the new domains and the tasks might be drastically different; (2) there might be very limited amount of annotated data on the new domain and (3) full training of a new model for each new task is prohibitive in terms of computation and memory, due to the sheer number of parameters of deep CNNs.
1 code implementation • 11 Apr 2019 • Adrian Bulat, Georgios Tzimiropoulos, Jean Kossaifi, Maja Pantic
Big neural networks trained on large datasets have advanced the state-of-the-art for a large variety of challenging problems, improving performance by a large margin.
no code implementations • CVPR 2019 • Jean Kossaifi, Adrian Bulat, Georgios Tzimiropoulos, Maja Pantic
In this paper, we propose to fully parametrize Convolutional Neural Networks (CNNs) with a single high-order, low-rank tensor.
Ranked #35 on Pose Estimation on MPII Human Pose
no code implementations • 27 Feb 2019 • Arinbjörn Kolbeinsson, Jean Kossaifi, Yannis Panagakis, Adrian Bulat, Anima Anandkumar, Ioanna Tzoulaki, Paul Matthews
CNNs achieve remarkable performance by leveraging deep, over-parametrized architectures, trained on large datasets.
1 code implementation • 14 Aug 2018 • Adrian Bulat, Georgios Tzimiropoulos
To this end, we make the following contributions: (a) we are the first to study the effect of neural network binarization on localization tasks, namely human pose estimation and face alignment.
Ranked #1 on 3D Face Alignment on AFLW2000-3D
2 code implementations • ECCV 2018 • Adrian Bulat, Jing Yang, Georgios Tzimiropoulos
This paper is on image and face super-resolution.
no code implementations • CVPR 2018 • Adrian Bulat, Georgios Tzimiropoulos
This paper addresses 2 challenging tasks: improving the quality of low resolution facial images and accurately locating the facial landmarks on such poor resolution images.
Ranked #4 on Face Hallucination on FFHQ 512 x 512 - 16x upscaling
1 code implementation • ICCV 2017 • Aaron S. Jackson, Adrian Bulat, Vasileios Argyriou, Georgios Tzimiropoulos
Our CNN works with just a single 2D facial image, does not require accurate alignment nor establishes dense correspondence between images, works for arbitrary facial poses and expressions, and can be used to reconstruct the whole 3D facial geometry (including the non-visible parts of the face) bypassing the construction (during training) and fitting (during testing) of a 3D Morphable Model.
Ranked #2 on 3D Face Reconstruction on Florence
8 code implementations • ICCV 2017 • Adrian Bulat, Georgios Tzimiropoulos
To this end, we make the following 5 contributions: (a) we construct, for the first time, a very strong baseline by combining a state-of-the-art architecture for landmark localization with a state-of-the-art residual block, train it on a very large yet synthetically expanded 2D facial landmark dataset and finally evaluate it on all other 2D facial landmark datasets.
Ranked #1 on Face Alignment on LS3D-W Balanced
3 code implementations • ICCV 2017 • Adrian Bulat, Georgios Tzimiropoulos
(d) We present results for experiments on the most challenging datasets for human pose estimation and face alignment, reporting in many cases state-of-the-art performance.
Ranked #1 on Face Alignment on AFLW-Full
1 code implementation • 29 Sep 2016 • Adrian Bulat, Georgios Tzimiropoulos
This paper describes our submission to the 1st 3D Face Alignment in the Wild (3DFAW) Challenge.
Ranked #1 on Face Alignment on 3DFAW
no code implementations • British Machine Vision Conference 2016 • Adrian Bulat, Georgios Tzimiropoulos
Besides playing the role of a graphical model, CNN regression is a key feature of our system, guiding the network to rely on context for predicting the location of occluded landmarks, typically encountered in very large poses.
Ranked #1 on Face Alignment on AFLW-PIFA (21 points)
1 code implementation • 6 Sep 2016 • Adrian Bulat, Georgios Tzimiropoulos
Our main contribution is a CNN cascaded architecture specifically designed for learning part relationships and spatial context, and robustly inferring pose even for the case of severe part occlusions.
Ranked #10 on Pose Estimation on Leeds Sports Poses