this article presents a study on facial emotion detection using deep learning, focusing on the challenges of emotion recognition in uncontrolled environments. the authors propose a multitask learning approach to improve the performance of facial emotion recognition by sharing a common feature representation with other related tasks, such as facial action unit detection. the proposed method uses a convolutional neural network (cnn) for training on a public face database, extracting seven probabilities for each frame of the face, aggregating single-frame probabilities into fixed-length image descriptors, and classifying all images using a support vector machine (svm) trained on image descriptors. the study validates the proposal using three datasets acquired in non-controlled environments and an application to predict compound facial emotion expressions. the authors highlight the importance of feature engineering in improving the performance of facial emotion recognition systems, noting that traditional approaches often rely on edge information due to its variability in individual expressions. the study also discusses the need for high-performance computing resources, such as a high-end graphics processing unit (gpu), to process large quantities of test data efficiently. the authors conclude that the proposed system has the potential to significantly improve the accuracy of facial emotion recognition, with the possibility of achieving over 97% accuracy when sufficient data and high-performance computing resources are available. the study emphasizes the importance of addressing the challenges of emotion recognition in uncontrolled environments and the potential of deep learning in this area.this article presents a study on facial emotion detection using deep learning, focusing on the challenges of emotion recognition in uncontrolled environments. the authors propose a multitask learning approach to improve the performance of facial emotion recognition by sharing a common feature representation with other related tasks, such as facial action unit detection. the proposed method uses a convolutional neural network (cnn) for training on a public face database, extracting seven probabilities for each frame of the face, aggregating single-frame probabilities into fixed-length image descriptors, and classifying all images using a support vector machine (svm) trained on image descriptors. the study validates the proposal using three datasets acquired in non-controlled environments and an application to predict compound facial emotion expressions. the authors highlight the importance of feature engineering in improving the performance of facial emotion recognition systems, noting that traditional approaches often rely on edge information due to its variability in individual expressions. the study also discusses the need for high-performance computing resources, such as a high-end graphics processing unit (gpu), to process large quantities of test data efficiently. the authors conclude that the proposed system has the potential to significantly improve the accuracy of facial emotion recognition, with the possibility of achieving over 97% accuracy when sufficient data and high-performance computing resources are available. the study emphasizes the importance of addressing the challenges of emotion recognition in uncontrolled environments and the potential of deep learning in this area.