arXiv:2102.11361v3 Announce Type: replace 
Abstract: FaCells is a method, and an exhibition, that turns model internals into line based artworks. Aligned face photographs (CelebA, 260k images, 40 attributes) are translated into vector sketches suitable for an XY plotter. We study how to 'write' these drawings for a sequence model, comparing absolute vs. relative point encodings and random vs. travel-minimizing stroke order. A bidirectional LSTM is trained for attribute prediction; a minimal architectural change, removing the global average over the sequence and applying a Dense layer at each point, yields per point attribute scores. Aggregating points whose score exceeds an attribute specific threshold across many portraits produces new drawings we call FaCells: statistical abstractions of attributes such as Eyeglasses, Wavy Hair, or Bangs. Across ablations, absolute coordinates with travel-minimizing order and a global average readout perform best; this configuration is then adapted to produce per-point scores. Multilabel training over 40 attributes is stable, and attributes reaching at least 50% balanced accuracy are visualized as FaCells. Complementary notions (e.g., No_Beard) are constructed by selecting points below a negative threshold. FaCells foregrounds interpretability as a creative tool: the resulting works are plotter ready, reproducible, and inexpensive to realize, yet materially present. Presented at Spectrum Miami 2025, the project bridges data, model, and paper while acknowledging the limits of the labels and the biases of the dataset.

FaCells هي طريقة مبتكرة تحول النماذج الداخلية إلى أعمال فنية قائمة على الخطوط، باستخدام صور الوجوه المتوافقة من مجموعة بيانات CelebA لإنشاء رسومات متجهة مناسبة للطابعات XY. تركز الدراسة على تحسين ترميزات النقاط وترتيب الضربات لنموذج LSTM ثنائي الاتجاه تم تدريبه على التنبؤ بسمات الوجه، مما يؤدي إلى درجات سمات لكل نقطة تؤدي إلى تمثيلات فنية جديدة تُسمى FaCells.

FaCells es un método innovador que transforma los internos del modelo en obras de arte basadas en líneas, utilizando fotografías de rostros alineadas del conjunto de datos CelebA para crear bocetos vectoriales para plotters XY. El estudio se centra en optimizar las codificaciones de puntos y el orden de los trazos para un LSTM bidireccional entrenado para predecir atributos faciales, lo que resulta en puntuaciones de atributos por punto que conducen a nuevas representaciones artísticas llamadas FaCells.

FaCells est une méthode novatrice qui transforme les internes de modèle en œuvres d'art basées sur des lignes, en utilisant des photographies de visages alignées du jeu de données CelebA pour créer des croquis vectoriels pour des traceurs XY. L'étude se concentre sur l'optimisation des encodages de points et de l'ordre des coups pour un LSTM bidirectionnel formé à prédire les attributs faciaux, aboutissant à des scores d'attributs par point qui donnent lieu à de nouvelles représentations artistiques appelées FaCells.

FaCells is a novel method that transforms model internals into line-based artworks, utilizing aligned face photographs from the CelebA dataset to create vector sketches for XY plotters. The study focuses on optimizing point encodings and stroke order for a bidirectional LSTM trained to predict facial attributes, resulting in per-point attribute scores that lead to new artistic representations called FaCells.

FaCells. Teaching Machines the Language of Lines: Per Point Attribute Scores for Face-Sketch Classification

arXiv:1809.01804v3 Announce Type: replace-cross 
Abstract: In the field of machine learning, it is still a critical issue to identify and supervise the learned representation without manually intervening or intuition assistance to extract useful knowledge or serve for the downstream tasks. In this work, we focus on supervising the influential factors extracted by the variational autoencoder(VAE). The VAE is proposed to learn independent low dimension representation while facing the problem that sometimes pre-set factors are ignored. We argue that the mutual information of the input and each learned factor of the representation plays a necessary indicator of discovering the influential factors. We find the VAE objective inclines to induce mutual information sparsity in factor dimension over the data intrinsic dimension and therefore result in some non-influential factors whose function on data reconstruction could be ignored. We show mutual information also influences the lower bound of the VAE's reconstruction error and downstream classification task. To make such indicator applicable, we design an algorithm for calculating the mutual information for the VAE and prove its consistency. Experimental results on MNIST, CelebA and DEAP datasets show that mutual information can help determine influential factors, of which some are interpretable and can be used to further generation and classification tasks, and help discover the variant that connects with emotion on DEAP dataset.

تركزت دراسة حديثة على العوامل المؤثرة المستخرجة من المشفرات التلقائية المتغيرة (VAEs)، مما يبرز التحدي المتمثل في الإشراف على التمثيلات المتعلمة دون تدخل يدوي. تؤكد الأبحاث على دور المعلومات المتبادلة بين المدخلات والعوامل المتعلمة كدليل رئيسي لتحديد العوامل المؤثرة، مما يكشف أن بعض العوامل قد تكون غير مؤثرة ويمكن تجاهلها في إعادة بناء البيانات.

Un estudio reciente se ha centrado en los factores influyentes extraídos por los autoencoders variacionales (VAE), destacando el desafío de supervisar las representaciones aprendidas sin intervención manual. La investigación enfatiza el papel de la información mutua entre las entradas y los factores aprendidos como un indicador clave para identificar factores influyentes, revelando que algunos factores pueden ser no influyentes y pueden ser ignorados en la reconstrucción de datos.

Une étude récente s'est concentrée sur les facteurs influents extraits par les autoencodeurs variationnels (VAE), soulignant le défi de superviser les représentations apprises sans intervention manuelle. La recherche met en avant le rôle de l'information mutuelle entre les entrées et les facteurs appris comme un indicateur clé pour identifier les facteurs influents, révélant que certains facteurs peuvent être non influents et peuvent être négligés dans la reconstruction des données.

A recent study has focused on the influential factors extracted by variational autoencoders (VAEs), highlighting the challenge of supervising learned representations without manual intervention. The research emphasizes the role of mutual information between inputs and learned factors as a key indicator for identifying influential factors, revealing that some factors may be non-influential and can be disregarded in data reconstruction.

FaCells. Teaching Machines the Language of Lines: Per Point Attribute Scores for Face-Sketch Classification

Was this article worth reading? Share it

Videolulu

AiReelGenerator.com

Fakeface