We present an end-to-end trainable multi-task model that locates and retrieves target objects from multi-object scenes. The model is an extension of the Siamese Mask R-CNN, which combines the components of Siamese Neural Networks (SNNs) and Mask R-CNN for performing one-shot instance segmentation. The proposed network, called Grasping Siamese Mask R-CNN (GSMR-CNN), extends Siamese Mask R-CNN by adding an additional branch for grasp detection in parallel to the previous object detection head branches. This allows our model to identify a target object with a suitable grasp simultaneously, as opposed to other approaches that require the training of separate models to achieve the same task. The inherent SNN properties enable the proposed model to generalize and recognize new object categories that were not present during training, which is beyond the capabilities of standard object detectors. Moreover, an end-to-end solution uses shared features entailing less model parameters. The model achieves grasp accuracy scores of 92.1 % and 90.4% on the OCID grasp dataset on image-wise and object-wise splits. Physical experiments show that the model achieves a grasp success rate of 76.4 % when correctly identifying the object. Code and models are available at https://github.com/valerijah/grasping_siamese_mask_rcnn

The effective modelling of high-dimensional data with hundreds to thousands of features remains a challenging task in the field of machine learning. This process is a manually intensive task and requires skilled data scientists to apply exploratory data analysis techniques and statistical methods in pre-processing datasets for meaningful analysis with machine learning methods. However, the massive growth of data has brought about the need for fully automated data analysis methods. One of the key challenges is the accurate selection of a set of relevant features, which can be buried in high-dimensional data along with irrelevant noisy features, by choosing a subset of the complete set of input features that predicts the output with higher accuracy comparable to the performance of the complete input set. Kohonen's self-organising neural network map has been utilised in various ways for this task, such as with the weighted self-organising map (WSOM) approach and this method is reviewed for its efficacy. The study demonstrates that the WSOM approach can result in different results on different runs on a given dataset due to the inappropriate use of the steepest descent optimisation method to minimise the weighted SOM's cost function. An alternative feature weighting approach based on analysis of the SOM after training is presented; the proposed approach allows the SOM to converge before analysing the input relevance, unlike the WSOM that aims to apply weighting to the inputs during the training which distorts the SOM's cost function, resulting in multiple local minimums meaning the SOM does not consistently converge to the same state. We demonstrate the superiority of the proposed method over the WSOM and a standard SOM in feature selection with improved clustering analysis.