Milan Sulc

Rahaf Aljundi, Yash Patel, Milan Sulc et al.

Cross entropy loss has served as the main objective function for classification-based tasks. Widely deployed for learning neural network classifiers, it shows both effectiveness and a probabilistic interpretation. Recently, after the success of self supervised contrastive representation learning methods, supervised contrastive methods have been proposed to learn representations and have shown superior and more robust performance, compared to solely training with cross entropy loss. However, cross entropy loss is still needed to train the final classification layer. In this work, we investigate the possibility of learning both the representation and the classifier using one objective function that combines the robustness of contrastive learning and the probabilistic interpretation of cross entropy loss. First, we revisit a previously proposed contrastive-based objective function that approximates cross entropy loss and present a simple extension to learn the classifier jointly. Second, we propose a new version of the supervised contrastive training that learns jointly the parameters of the classifier and the backbone of the network. We empirically show that our proposed objective functions show a significant improvement over the standard cross entropy loss with more training stability and robustness in various challenging settings.

Tomas Sipka, Milan Sulc, Jiri Matas

In many computer vision classification tasks, class priors at test time often differ from priors on the training set. In the case of such prior shift, classifiers must be adapted correspondingly to maintain close to optimal performance. This paper analyzes methods for adaptation of probabilistic classifiers to new priors and for estimating new priors on an unlabeled test set. We propose a novel method to address a known issue of prior estimation methods based on confusion matrices, where inconsistent estimates of decision probabilities and confusion matrices lead to negative values in the estimated priors. Experiments on fine-grained image classification datasets provide insight into the best practice of prior shift estimation and classifier adaptation, and show that the proposed method achieves state-of-the-art results in prior adaptation. Applying the best practice to two tasks with naturally imbalanced priors, learning from web-crawled images and plant species classification, increased the recognition accuracy by 1.1% and 3.4% respectively.

Milan Sulc, Jiri Matas

The problem of different training and test set class priors is addressed in the context of CNN classifiers. We compare two different approaches to estimating the new priors: an existing Maximum Likelihood Estimation approach (optimized by an EM algorithm or by projected gradient descend) and a proposed Maximum a Posteriori approach, which increases the stability of the estimate by introducing a Dirichlet hyper-prior on the class prior probabilities. Experimental results show a significant improvement on the fine-grained classification tasks using known evaluation-time priors, increasing the top-1 accuracy by 4.0% on the FGVC iNaturalist 2018 validation set and by 3.9% on the FGVCx Fungi 2018 validation set. Estimation of the unknown test set priors noticeably increases the accuracy on the PlantCLEF dataset, allowing a single CNN model to achieve state-of-the-art results and outperform the competition-winning ensemble of 12 CNNs. The proposed Maximum a Posteriori estimation increases the prediction accuracy by 2.8% on PlantCLEF 2017 and by 1.8% on FGVCx Fungi, where the existing MLE method would lead to a decrease accuracy.