Бидний тухай
Багш ажилтан
Хураангуй—“Эрдэнэт үйлдвэр” ТӨҮГ -ын Ил уурхайн гол малталтын төхөөрөмж бол шанага юм. Малталтын явцад шанаганы шүд хүдэрт удаан хугацаагаар шууд нөлөөлснөөр шанаганы шүд нь хүчтэй цохилт, үрэлт, бусад хүчинд өртөж, улмаар шанаганы шүд нь шүдний сууринаас гэнэт сулардаг, зарим тохиолдолд бүр хугардаг ба тус шүдийг хүдрийн хамт бутлуурт буулгасны дараа энэ нь бутлуурыг эвдрэлд хүргэж, уурхайн бутлах үйлдвэрлэлийн шугамд бүхэлд нь нөлөөлдөг. эдийн засгийн, хүний болон материаллаг нөөцийн үр ашгийн алдагдал үүсэх эрсдэлтэй. Шанаганы шүд нь өндөр манганы ган эсвэл хайлштай гангаар хийгдсэн байдаг тул хатуулаг нь хүдэр болон бусад материалынхаас хамаагүй өндөр байдаг. Энэхүү судалгааны ажлаар Экскаваторын шанаганы дүрсийг цуглуулж өгөгдлийн багц бэлтгэн YOLOX болон YOLO-NAS загваруудыг сургаж харьцуулан шанаганы шүдийг илрүүлсэн. Мөн шанаган дээрх хүдрийн дүрсэд сегментчлэл хийдэг SAM загвар болон дүрсээс ирмэг илрүүлэх DexiNED загвар ашиглан бүхэллэгийн хэмжээг ойролцоогоор тодорхойлсон. Шүдний уналтыг илрүүлэх загварууд 87% -тай илрүүлсэн ба хүдрийн бүхэллэгийг SAM загвараар дүрсээс чулууг сегменчлэх замаар тодорхойлох илүү үр дүнтэй байна.
Pediatric pneumonia is the leading cause of morbidity and mortality among infectious diseases, but there is no reliable way to diagnose the disease other than through biochemical tests and X-rays. Due to changes in breath sounds during lung disease, it can be detected in the early stages using a stethoscope. However, hearing these lung changes with a traditional stethoscope is difficult when common symptoms are few and clinical symptoms are vague and subtle. Research on effective pneumonia disease recognition models using machine learning and deep learning methods has mainly been based on lung radiographs. Our research, however, was based on chest sounds obtained using an electronic stethoscope, which offers improved accessibility and ease of use. 170 chest sound recordings were collected from 35 healthy individuals, and 30 recordings from 2 individuals diagnosed with pneumonia were recorded using an electronic stethoscope. Since respiratory sounds in pneumonia are characterized by low-frequency crackles and high-frequency wheezes, we applied machine learning and deep learning algorithms considering the frequency spectrogram and the Mel-frequency cepstral coefficients as the main features of the audio data. Three types of deep learning methods were used: CNN, LSTM, and a combination of CNN and LSTM. The models were decisively trained for 100 epochs with a learning rate of 0.001 using the Adam optimizer. The CNN model took about 2 hours to achieve an optimal accuracy of 82.19%, while the LSTM model took about 7 hours and achieved 84.58% optimal accuracy. By combining CNN and LSTM, the training time was reduced to 3 hours, resulting in an optimal accuracy of 85.33% and an identification accuracy of pneumonia at 91%.
In recent years, many researchers have shown increasing interest in music information retrieval (MIR) applications, with automatic chord recognition being one of the popular tasks. Many studies have achieved/demonstrated considerable improvement using deep learning based models in automatic chord recognition problems. However, most of the existing models have focused on simple chord recognition, which classifies the root note with the major, minor, and seventh chords. Furthermore, in learning-based recognition, it is critical to collect high-quality and large amounts of training data to achieve the desired performance. In this paper, we present a multi-task learning (MTL) model for a guitar chord recognition task, where the model is trained using a relatively large-vocabulary guitar chord dataset. To solve data scarcity issues, a physical data augmentation method that directly records the chord dataset from a robotic performer is employed. Deep learning based MTL is proposed to improve the performance of automatic chord recognition with the proposed physical data augmentation dataset. The proposed MTL model is compared with four baseline models and its corresponding single-task learning model using two types of datasets, including a human dataset and a human combined with the augmented dataset. The proposed methods outperform the baseline models, and the results show that most scores of the proposed multi-task learning model are better than those of the corresponding single-task learning model. The experimental results demonstrate that physical data augmentation is an effective method for increasing the dataset size for guitar chord recognition tasks.
In this paper, we propose a new framework for generating big sized dataset using synthetic data generation by robotics. In learning-based recognition, for example, using convolutional neural networks (CNNs), it is critical for the performance, to collect high-quality and large amounts of training data. Previously, to increase the training data set, a data augmentation technique based on digital signal processing were applied to the original sound data. However, the data augmentation based on digital signal processing data is a limited method, because it depends on some previous knowledge of the data and cannot perform for all domains. On the other hand, we propose a new dataset collection technique using a robot that automatically plays instruments, by which it becomes possible to add high-quality data to training samples. Experimental results for guitar chord recognition show that the proposed method using CNNs and a guitar robot can outperform the CNN systems with the traditional data augmentation.
