This paper presents three fully convolutional neural network (FCNN) architectures for change detection using pairs of coregistered images. The proposed methods are trained end-to-end from scratch using annotated change detection datasets. The architectures are extensions of previous work, with improvements in accuracy and inference speed. Two new fully convolutional Siamese architectures are introduced, using skip connections to enhance performance. The methods are tested on two open datasets: the Onera Satellite Change Detection (OSCD) dataset and the Air Change (AC) dataset. The results show that the proposed methods outperform previous approaches in terms of accuracy and speed, with inference times under 0.1 seconds per image. The FC-Siam-diff architecture achieves the best performance, followed by FC-EF. The methods are efficient and suitable for processing large-scale Earth observation data from systems like Copernicus and Landsat. The paper concludes that the proposed FCNNs are effective for change detection, with potential for further improvements through larger datasets and other image modalities.This paper presents three fully convolutional neural network (FCNN) architectures for change detection using pairs of coregistered images. The proposed methods are trained end-to-end from scratch using annotated change detection datasets. The architectures are extensions of previous work, with improvements in accuracy and inference speed. Two new fully convolutional Siamese architectures are introduced, using skip connections to enhance performance. The methods are tested on two open datasets: the Onera Satellite Change Detection (OSCD) dataset and the Air Change (AC) dataset. The results show that the proposed methods outperform previous approaches in terms of accuracy and speed, with inference times under 0.1 seconds per image. The FC-Siam-diff architecture achieves the best performance, followed by FC-EF. The methods are efficient and suitable for processing large-scale Earth observation data from systems like Copernicus and Landsat. The paper concludes that the proposed FCNNs are effective for change detection, with potential for further improvements through larger datasets and other image modalities.