H.5. Image Processing and Computer Vision
Pouria Maleki; Abbas Ramazani; Hassan Khotanlou; Sina Ojaghi
Abstract
Providing a dataset with a suitable volume and high accuracy for training deep neural networks is considered to be one of the basic requirements in that a suitable dataset in terms of the number and quality of images and labeling accuracy can have a great impact on the output accuracy of the trained ...
Read More
Providing a dataset with a suitable volume and high accuracy for training deep neural networks is considered to be one of the basic requirements in that a suitable dataset in terms of the number and quality of images and labeling accuracy can have a great impact on the output accuracy of the trained network. The dataset presented in this article contains 3000 images downloaded from online Iranian car sales companies, including Divar and Bama sites, which are manually labeled in three classes: car, truck, and bus. The labels are in the form of 5765 bounding boxes, which characterize the vehicles in the image with high accuracy, ultimately resulting in a unique dataset that is made available for public use.The YOLOv8s algorithm, trained on this dataset, achieves an impressive final precision of 91.7% for validation images. The Mean Average Precision (mAP) at a 50% threshold is recorded at 92.6%. This precision is considered suitable for city vehicle detection networks. Notably, when comparing the YOLOv8s algorithm trained with this dataset to YOLOv8s trained with the COCO dataset, there is a remarkable 10% increase in mAP at 50% and an approximately 22% improvement in the mAP range of 50% to 95%.
H.6.5.13. Signal processing
M. Asadolahzade Kermanshahi; M. M. Homayounpour
Abstract
Improving phoneme recognition has attracted the attention of many researchers due to its applications in various fields of speech processing. Recent research achievements show that using deep neural network (DNN) in speech recognition systems significantly improves the performance of these systems. There ...
Read More
Improving phoneme recognition has attracted the attention of many researchers due to its applications in various fields of speech processing. Recent research achievements show that using deep neural network (DNN) in speech recognition systems significantly improves the performance of these systems. There are two phases in DNN-based phoneme recognition systems including training and testing. Most previous research attempted to improve training phase such as training algorithms, different types of network, network architecture, feature type, etc. But in this study, we focus on test phase which is related to generate phoneme sequence that is also essential to achieve good phoneme recognition accuracy. Past research used Viterbi algorithm on hidden Markov model (HMM) to generate phoneme sequences. We address an important problem associated with this method. To deal with the problem of considering geometric distribution of state duration in HMM, we use real duration probability distribution for each phoneme with the aid of hidden semi-Markov model (HSMM). We also represent each phoneme with only one state to simply use phonemes duration information in HSMM. Furthermore, we investigate the performance of a post-processing method, which corrects the phoneme sequence obtained from the neural network, based on our knowledge about phonemes. The experimental results using the Persian FarsDat corpus show that using extended Viterbi algorithm on HSMM achieves phoneme recognition accuracy improvements of 2.68% and 0.56% over conventional methods using Gaussian mixture model-hidden Markov models (GMM-HMMs) and Viterbi on HMM, respectively. The post-processing method also increases the accuracy compared to before its application.
