Abstract: A method for determining the presence of at least one defect in a three-dimensional part on the basis of a reference base, wherein: an image of the part is acquired by tomography; the image is segmented and the segmented image is registered with respect to the images of the reference base; the subdomains associated with the registered segmented image are determined; and for each subdomain of the registered segmented image, the following steps are carried out: registering the subdomain; breaking down the subdomain with respect to the principal modes of the reference base; generating a synthetic image of the subdomain from the breakdown, which is subtracted from the registered segmented image of the subdomain; and determining that a defect is present on the basis of the comparison of the value of each pixel of the residual image with a predetermined threshold.

Abstract: A method for detecting defects during laser additive manufacturing, comprising the following steps: B21) a first image is captured, the first image being the image, captured in the infrared spectrum, of an upper surface of a layer of powder exposed to laser scan; B23) the first image is processed using a first convolutional neural network of self-encoding type, in such a way as to produce a defect mask indicating the location of defects at the upper surface of the layer of powder. A method for manufacturing parts during which the presence of defects is detected using the preceding method. A data-processing device, computer program and storage medium for the implementation of this method.