High calorie intake has proved harmful worldwide, as it has led to many diseases. However, dieticians have deemed that a standard intake of number of calories is essential to maintain the right balance of calorie content in human body. In this thesis, we consider the category of tools that use image processing to recognize single and multiple mixed-food objects, namely Deep Learning and the Support Vector Machine (SVM). We propose a method for the fully automatic and user-friendly calibration of the sizes of food portions. This calibration is required to estimate the total number of calories in food portions. In this work, to compute the number of calories in the food object, we go beyond the finger-based calorie calibration method that has been used in the past, by automatically measuring the distance between the user and the food object. We implement a block resize method that uses the measured distance values along with the recognized food object name to further estimate calories. While measuring distance, the system also assists the user in real time to capture an image that enables the quick and accurate calculation of the number of calories in the food object. The experimental results showed that our method, which uses deep learning to analyze food objects, led to an improvement of 16.58% in terms of recognition, over the SVM-based method. Moreover, the block resize method showed that percentage error for calorie estimation was reduced to 3.64% as compared to 5% achieved in previous methods.

Les labels présents sur les emballages (e.g., « bio », « commerce équitable ») peuvent pousser les individus à sous-évaluer le contenu calorique des aliments, créant ainsi un effet de halo santé. Dans cette thèse, nous défendons l’idée qu’un mécanisme de fausse attribution affective pourrait en partie expliquer cet effet de halo santé. Nous présenterons 11 études ayant pour but de tester empiriquement cette hypothèse. Parmi les principaux résultats, nous avons montré qu’il était possible d’observer cet effet même si les individus ne rapportent pas avoir pris en compte le label dans leur évaluation (Etudes 3) et dans un contexte de double tâche entravant un raisonnement délibéré de leur part (Etudes 4 et 5). Nous avons également pu montrer, par le biais d’une tâche de fausse attribution affective, qu’un stimulus neutre, lorsqu’il est précédé d’un label bio (vs. une image contrôle) est ensuite évalué plus positivement (Etudes 6, 7 et 8). Enfin, nous avons observé une congruence systématique entre la valence de l’amorce (i.e., le label) et la valence de l’évaluation subséquente d’un stimulus neutre (i.e., le contenu calorique d’un produit alimentaire ; Etudes 9, 10 et 11). Ce biais cognitif relatif à nos évaluations caloriques apparaît donc comme un phénomène robuste, ne faisant intervenir que peu d’inférences délibérées de la part des individus et étant guidé par la valence du label (qu’il soit positif ou négatif). / Labels on food products can lead to unwarranted inferences: organic and fair-trade products are perceived as containing fewer calories. This effect is described in the literature as the health halo effect. In this thesis, we argue that an affective misattribution mechanism can partially explain this effect. We present 11 studies testing empirically this hypothesis.Among our main results, we show that this effect still occurs even if participants did not mention that they used the label in their evaluation (Study 3) and when they are under cognitive load, hindering a deliberate reasoning (Studies 4 and 5). By using an affective misattribution procedure, we also show that when a neutral stimulus is preceded by an organic label (vs. a control picture), this stimulus is then evaluated more positively (Studies 6, 7, and 8). Finally, we observed a systematic congruency between the valence of the prime (i.e., the label) and the valence of the evaluation of a neutral stimulus (i.e., the caloric content of a food product; Studies 9, 10, and 11). Taken together, these studies suggest that this cognitive bias is a robust phenomenon, involving a few inferences and mainly driven by the valence of the label (whether positive or negative).