Machine learning technique for morphological classification of galaxies from SDSS. IV. CNN-assisted expert validation of morphological catalogues for merging, irregular, edge-on, barred, ringed, and dust-lane galaxies at 0.003 < z < 0.1

Convolutional neural networks (CNNs) have become an efficient tool for automated classification of galaxy morphology in large astronomical surveys. However, identifying complex morphological features remains limited by projection effects, image artefacts, and intrinsic ambiguities, making expert visual validation essential to building reliable morphological catalogues. This paper introduces a CNN-assisted expert validation framework for constructing reliable galaxy image-based morphological catalogues. Our goal was to use this framework to perform a comprehensive visual validation of Sloan Digital Sky Survey (SDSS) galaxies at 0.003<z<0.1 previously classified by a CNN as merging, irregular, edge-on, barred, ringed, or dust-lane galaxies; to quantify the reliability and failure modes of CNN classifications; to construct homogeneous and visually verified catalogues of these morphological classes; and to characterise the central ionisation of the verified galaxies with emission-line diagnostics. We visually inspected all CNN-selected candidates in four morphological categories, namely merging (2,574), irregular (9,432), ringed (13,882), and dust-lane (588) galaxies, covering the full range of their catalogue feature scores. For edge-on and barred galaxies, we inspected the 17,000 and 4,323 highest-scoring candidates, respectively, which we selected from larger CNN catalogues. We cross-matched the refined samples with the Galaxy Zoo 2 (GZ2) catalogue, and we visually classified galaxies outside its selection domain, which carry no GZ2 morphological label. For the subset with SDSS spectroscopy, we applied the Baldwin--Phillips--Terlevich (BPT) and the N, ii /Hα versus Hα equivalent width (WHAN) diagnostics to the MPA--JHU emission-line measurements. We present homogeneous, visually verified catalogues containing 612 merging galaxies, 9,372 irregular galaxies, 16,822 edge-on galaxies, 811 barred galaxies, 2,145 ringed galaxies, and 575 dust-lane galaxies. Comparison with GZ2 demonstrates that CNN-assisted expert validation both refines existing classifications and extends reliable morphological catalogues beyond the effective Galaxy Zoo selection domain. The principal causes of CNN misclassification are projection effects, foreground stars, image artefacts, faint tidal structures, and intrinsically ambiguous morphologies. Emission-line diagnostics show that 87.8% of the ringed galaxies with low-ionisation nuclear emission-line region (LINER)-like line ratios have Hα equivalent widths consistent with ionisation by evolved stars rather than by an active nucleus. We also identified five strong new candidates for polar ring galaxies. Our results demonstrate that CNN-assisted expert validation remains essential for constructing reliable galaxy morphological catalogues. The resulting homogeneous datasets provide valuable resources for studies of galaxy morphology, secular evolution, and nuclear activity, and they constitute high-quality benchmark samples for future machine learning applications and forthcoming wide-field surveys.