Mask RCNN

Asset owners are required by the government to carry out regular inspection surveys to ensure the integrity of all pressure-containing equipment. Conventionally, such operations are performed manually by teams of trained inspectors through visual examinations on-site or remotely. We present an integrated framework for automating the entire process of pipeline inspection without any need for human intervention, pipeline function interruption, or equipment destruction by unleashing the power of the state-of-the-art digital technologies, including deep learning, computer vision, and cloud storage and computing.
Decades of survey videos captured by remotely operated vehicles (ROVs) and drones are broken into frames with optical character recognition (OCR)-extracted time/location stamps, which are annotated by our inspectors to precisely delineate the boundaries of the damaged areas and their specific categories. After extracting images from videos, augmenting the image data, the training set is fed into a deep convolutional neural network architecture equipped with instance segmentation layers for object detection. Trained models are then deployed on the cloud acting as intelligent inspectors for future surveys, allowing also to balance the trade-off between the inference accuracy and performance speed, being crucial to real time usage of the software.
After a pipe segment subject to damage is identified to be imposing an integrity risk, it will automatically be raised as a flag into our linked maintenance infrastructure along with the corresponding spatiotemporal information of the event to take the necessary maintenance, repair, or replacement actions.
The proposed endeavor outperforms common industry practices, as it not only reduces the asset operational costs by eliminating the need for the labor-intensive manual diagnostic inspections, but also improves the hazard mitigation plan by providing accurate risk assessments in shorter time spans.

Spotting the difference between bird species is a challenging task because the minute characteristics cannot be distinguished by the human eye. This identification problem is an example of fine grained classification task. The algorithm in such cases should be strong enough to identify small differences .However; there can be multiple problems such as orientation of bird, background colours, brightness and intensity of image which makes it difficult to correctly classify an image. The proposed paper presents a novel approach towards identification of species by increasing the resolution of image and applying Mask RCNN and custom net for classification. SRCNN with Mask RCNN and Inception V3 achieves 51.73% precison , 49.68% recall and 53.52 F1 Scor

In this study, chrome coatings were carried out using a Cr-III electroplating bath. The coated parts were classified depending on their appearance. A new approach was developed to classify the coated parts automatically using artificial intelligence methods. Mask RCNN and machine learning (ML) methods such as Multilayer Perceptron (MLP), Support Vector Classifier (SVC), Gaussian Process (GP), K-nearest Neighbors (KNN), XGBoost, and Random Forest Classifier (RFC) were used together. Mask RCNN was used to clean the coated parts from the redundant data. The extracted data were flattened and converted to the row vectors for use as input in ML methods. ML algorithms were used to classify the coated parts as “Pass” and “Fail”. The classification accuracy was checked with the leave one out (loo) cross-validation method. RFC method gave the highest accuracy, 0.83, and F1 score, 0.88. The accuracy of Mask RCNN was checked using a dataset of separated validation images. It was observed that extracting the unnecessary data from the images increased the accuracy exceedingly. Moreover, the method exhibits a high potential to keep the parameters of the electroplating process under control.