Academics discover black people'significantly more likely' to be identified when compared with other ethnic groups Essex police have paused the use of live facial recognition (LFR) technology after a study found cameras were significantly more likely to target black people than people of other ethnicities. The move to suspend use of the AI-enabled systems was revealed by the Information Commissioner's Office (ICO), which regulates the use of the technology deployed so far by at least 13 police forces in London, south and north Wales, Leicestershire, Northamptonshire, Hampshire, Bedfordshire, Suffolk, Greater Manchester, West Yorkshire, Surrey and Sussex. The ICO said Essex police had paused LFR deployments "after identifying potential accuracy and bias risks" and warned other forces to have mitigations in place. LFR systems are either mounted to fixed locations or deployed in vans. In January, the home secretary, Shabana Mahmood, announced the number of LFR vans would increase five-fold, with 50 available to every police force in England and Wales. Essex commissioned University of Cambridge academics to conduct a study, which involved 188 actors walking past cameras being actively deployed from marked police vans in Chelmsford.

Synthesizing realistic profile faces is promising for more efficiently training deep pose-invariant models for large-scale unconstrained face recognition, by populating samples with extreme poses and avoiding tedious annotations. However, learning from synthetic faces may not achieve the desired performance due to the discrepancy between distributions of the synthetic and real face images. To narrow this gap, we propose a Dual-Agent Generative Adversarial Network (DA-GAN) model, which can improve the realism of a face simulator's output using unlabeled real faces, while preserving the identity information during the realism refinement. The dual agents are specifically designed for distinguishing real v.s.

This work is motivated by the engineering task of achieving a near state-of-the-art face recognition on a minimal computing budget running on an embedded system. Our main technical contribution centers around a novel training method, called Multibatch, for similarity learning, i.e., for the task of generating an invariant ``face signature'' through training pairs of ``same'' and ``not-same'' face images. The Multibatch method first generates signatures for a mini-batch of $k$ face images and then constructs an unbiased estimate of the full gradient by relying on all $k^2-k$ pairs from the mini-batch. We prove that the variance of the Multibatch estimator is bounded by $O(1/k^2)$, under some mild conditions. In contrast, the standard gradient estimator that relies on random $k/2$ pairs has a variance of order $1/k$. The smaller variance of the Multibatch estimator significantly speeds up the convergence rate of stochastic gradient descent. Using the Multibatch method we train a deep convolutional neural network that achieves an accuracy of $98.2\%$ on the LFW benchmark, while its prediction runtime takes only $30$msec on a single ARM Cortex A9 core. Furthermore, the entire training process took only 12 hours on a single Titan X GPU.

Recognizing facial action units (AUs) from spontaneous facial expressions is still a challenging problem. Most recently, CNNs have shown promise on facial AU recognition. However, the learned CNNs are often overfitted and do not generalize well to unseen subjects due to limited AU-coded training images. We proposed a novel Incremental Boosting CNN (IB-CNN) to integrate boosting into the CNN via an incremental boosting layer that selects discriminative neurons from the lower layer and is incrementally updated on successive mini-batches. In addition, a novel loss function that accounts for errors from both the incremental boosted classifier and individual weak classifiers was proposed to fine-tune the IB-CNN. Experimental results on four benchmark AU databases have demonstrated that the IB-CNN yields significant improvement over the traditional CNN and the boosting CNN without incremental learning, as well as outperforming the state-of-the-art CNN-based methods in AU recognition. The improvement is more impressive for the AUs that have the lowest frequencies in the databases.

Your Out-of-Distribution Detection Method is Not Robust!

This protocol is often regarded as a "golden compass" for navigating SoT A detectors. But can these stand-out "winners" be

Existing FER methods typically report overall accuracy on highly imbalanced test sets but exhibit low performance in terms of the mean accuracy across all expression classes.

Synthesizing high-fidelity head avatars is a central problem for computer vision and graphics. While head avatar synthesis algorithms have advanced rapidly, the best ones still face great obstacles in real-world scenarios.