An Empirical Study on Ensemble of Segmentation Approaches

Abstract

Recognizing objects in images requires complex skills that involve knowledge about the context and the ability to identify the borders of the objects. In computer vision, this task is called semantic segmentation and it pertains to the classification of each pixel in an image. The task is of main importance in many real-life scenarios: in autonomous vehicles, it allows the identification of objects surrounding the vehicle; in medical diagnosis, it improves the ability of early detecting of dangerous pathologies and thus mitigates the risk of serious consequences. In this work, we propose a new ensemble method able to solve the semantic segmentation task. The model is based on convolutional neural networks (CNNs) and transformers. An ensemble uses many different models whose predictions are aggregated to form the output of the ensemble system. The performance and quality of the ensemble prediction are strongly connected with some factors; one of the most important is the diversity among individual models. In our approach, this is enforced by adopting different loss functions and testing different data augmentations. We developed the proposed method by combining DeepLabV3+, HarDNet-MSEG, and Pyramid Vision Transformers. The developed solution was then assessed through an extensive empirical evaluation in five different scenarios: polyp detection, skin detection, leukocytes recognition, environmental microorganism detection, and butterfly recognition. The model provides state-of-the-art results.

A Standardized Approach for Skin Detection: Analysis of the Literature and Case-Studies

“…In this section, we report the results about the empirical analysis performed for the hand segmentation task. We also provide an ablation study that shows the importance of adopting an ensemble based on DeepLabV3+; this ablation study, for the skin segmentation, have been already reported in [60].…”

“…• In [1] and [60] several deep learning segmentation approaches are compared, SegNet, U-Net, DeepLabv3+, HarD-NetMSEG (Harmonic Densely Connected Network) 1 [61] and Polyp-PVT [62] a deep learning segmentation model based on a transformer encoder, i.e. PVT (Pyramid Vision Transformer) 2 .…”

“…This loss has very good performances as it does not require neither any pre-or post-processing of the image nor a particular net in order to work. In [60] it is proposed to combine it with Dice Loss and weighted cross-entropy loss. • Contour-aware Loss was proposed for the first time in [88].…”

“…We applied these techniques to the training set on both input samples and masks. We adopt the two data augmentation techniques defined in [60], in particular:…”

Deep Semantic Segmentation in Skin Detection

“…DA2 is a more articulated procedure. We refer to [22] for an exhaustive description of the two approaches. Some artificial images (DA2 approach) contain only background pixels, to discard them we simply delete all the images where there are less than 10 pixels that belong to the foreground class.…”