Low Quality Video Research Articles

Deep Online Video Stabilization Using IMU Sensors

In this paper, we propose a deep learning based sensor-driven method for online video stabilization. This method utilizes euler angle and acceleration estimated from the gyroscope and accelerator to assist stable video reconstruction. We introduce two simple sub-networks for trajectory optimization, the first network exploits real unstable trajectory and acceleration of camera to detect shooting scenarios. This network also generates an attention mask to choose scenario-specific features adaptively. Then the second network predicts smooth camera paths based on real unstable trajectory using LSTM under the supervision of the former mask. The output of trajectory optimization network is filtered with a two-step modification process to guarantee the smoothness. The real and smoothed camera paths are then utilized as guidance to projectively generate stable frames. We also capture videos with sensor data covering seven typical shooting scenarios, and design a groundtruth generation method to construct pseud-labels. Moreover, the trajectory smoothing network allows 3 or 10 frames buffer as the future information to construct a lookahead filter. Experimental results show that our online method could outperform other offline state-of-the-art methods in several shaky video clips with fewer buffer frames over both general and low-quality videos. Furthermore, our method could effectively reduce running times without image content analysis, and the stabilization efficiency reaches 25 fps on 1080p videos.

Forgery detection of low quality deepfake videos

The rapid growth of online media over different social media platforms or over the internet along with many benefits have some negative effects as well. Deep learning has many positive applications like medical, animations and cybersecurity etc. But over the past few years, it is observed that it is been used for negative aspects as well such as defaming, black-mailing and creating privacy concerns for the general public. Deepfake is common terminology used for facial forgery of a person in media like images or videos.The advancement in the forgery creation area have challenged the researchers to create and develop advance forgery detection systems capable to detect facial forgeries. Proposed forgery detection system works on the CNN-LSTM model in which we first extracted faces from the frames using MTCNN then performed spatial feature extraction using pretrained Xception network and then used LSTM for temporal feature extraction. At the end classification is performed to predict the video as real or fake. The system is capable to detect low quality videos. The current system has shown good accuracy results for detecting real or fake videos on the Google deepfake AI dataset.

Examining the Technical Issues in Teletherapy for Children with Developmental Delays

Abstract. Medical digitization has advanced in developing countries, improving work efficiency. Teletherapy has emerged from this advancement. Nevertheless, teletherapy has encountered obstacles, such as software and hardware issues. Problems arose from technological factors such as slow connectivity, inadequate hardware specs, and poor audiovisual. This study aims to examine the technical issues related to teletherapy for children with developmental delays. By implementing descriptive statistical approaches with 43 participants, we found that patients struggled to be involved in teletherapy, resulting in the need for special guidance in the process. Therapists felt their clients' places were extremely crowded, and the low video quality of their devices hampered their ability to conduct optimal observations. Patients had a higher level of satisfaction and interest than therapists, meaning the patients would likely continue to be involved in the future. We also came up with another finding indicating that confidentiality has not become a major concern since building an intimate therapeutic relationship and learning to operate technology are still top priorities. Henceforth, implementing teletherapy would need a solid strategy in the future. Keywords: Children, digitization, technical issues, teletherapy, therapists

LiveSR: Enabling Universal HD Live Video Streaming With Crowdsourced Online Learning

The high-definition (HD) live video streaming has gained significant popularity due to the rapid growth of 4G/5G and social media. However, for devices with constrained bandwidth, they still have no sufficient bandwidth to support HD live video streaming. In this paper, we propose a neural-enhanced HD live video streaming framework called LiveSR to provide universal HD live video streaming for both bandwidth-constrained and bandwidth-rich devices. For bandwidth-constrained devices, LiveSR delivers low-quality video streams and then boosts video quality at the device side with super-resolution (SR) techniques. The difficulty lies in how to train the SR model with low cost and conduct quality enhancement in real time. To address these challenges, we design a crowdsourced online training method by exploiting computation resources and HD video data on bandwidth-rich devices in the same video channel. We also propose an imitation learning-based decision making algorithm to make downloading decisions for video chunks and SR models under limited bandwidth. We implement and evaluate our proposed LiveSR framework using real network traces, and the experiment results show that LiveSR outperforms all the other baseline approaches, with 65.5% improvement in terms of the average QoE and 5.7% in terms of video quality (i.e., PSNR), and the achieved frame rate can be as high as 30 frames per second.

Adaptive Weighted Flow Net Algorithm for Human Activity Recognition Using Depth Learned Features

Human Activity Recognition (HAR) from video data collections is the core application in vision tasks and has a variety of utilizations including object detection applications, video-based behavior monitoring, video classification, and indexing, patient monitoring, robotics, and behavior analysis. Although many techniques are available for HAR in video analysis tasks, most of them are not focusing on behavioral analysis. Hence, a new HAR system analysis the behavioral activity of a person based on the deep learning approach proposed in this work. The most essential aim of this work is to recognize the complex activities that are useful in many tasks that are based on object detection, modelling of individual frame characteristics, and communication among them. Moreover, this work focuses on finding out the human actions from various video resolutions, invariant human poses, and nearness of multi objects. First, we identify the key and essential frames of each activity using histogram differences. Secondly, Discrete Wavelet Transform (DWT) is used in this system to extract coefficients from the sequence of key-frames where the activity is localized in space. Finally, an Adaptive Weighted Flow Net (AWFN) algorithm is proposed in this work for effective video activity recognition. Moreover, the proposed algorithm has been evaluated by comparing it with the existing Visual Geometry Group (VGG-16) convolution neural networks for making performance comparisons. This work focuses on competent deep learning-based feature extraction to discriminate the activities for performing the classification accuracy. The proposed model has been evaluated with VGG-16 using a combination of regular UCF-101 activity datasets and also in very challenging Low-quality videos such as HMDB51. From these investigations, it is proved that the proposed AWFN approach gives higher detection accuracy of 96%. It is approximately 0.3% to 7.88% of higher accuracy than state-of-art methods.

A Model of the Centrifugal Pedagogical Process in a Digital Environment

The leading place in the pedagogical science of Russia and the CIS countries has long been occupied by the idea of reviving the humanistic tradition, but the educational process in schools of different types still follows traditional authoritarian technology, as it did many years ago, and digitalization is manifested by low-quality video tutorials, fictitious electronic magazine, etc. According to the authors, digitalization contributes to increased comfort due to gamification, the possibility of repetition and consolidation with the use of games, simulators. But foreign studies are now increasingly focusing on the consequences and risks of digitalization for humans. Due to the influence of the Internet and digital technologies on the education of students, the need to change the models of education and upbringing with the development of the readiness of teachers and students to interact in a digital environment is actualized. Goal. To develop a model of a system of centrifugal pedagogical process based on a personality-centered approach in education, when the source of development is the student himself, with the harmonization of psychological and pedagogical support for the process of formation of creative self-realization of the individual in a digital environment. Methodology and methods of research. For comparative analysis, the author has chosen a model of the system of the traditional educational process (according to P. I. Pidkasistom). The study was conducted on the basis of the personality-centered approach of K. Rogers, the concept of social education in the process of socialization (I. P. Ivanov, A. V. Mudrik, T. A. Romm), a systematic, synergetic approach (L. I. Novikova), methods of analysis, establishment of causal relationships, processing of bibliographic sources, experimental work. In conclusion, it is concluded that: in traditional schools, it is possible to transfer education and upbringing to a centrifugal systematic pedagogical technology of teaching and upbringing within the framework of compliance with the requirements of state educational standards with the implementation of individual activities on their own motivation; many methods that provide a stable educational effect, as a unique method of collective creative activity of I. P. Ivanov, today proved to be particularly relevant for the educational activity of the teacher.

No-Reference Video Quality Assessment Using the Temporal Statistics of Global and Local Image Features.

During acquisition, storage, and transmission, the quality of digital videos degrades significantly. Low-quality videos lead to the failure of many computer vision applications, such as object tracking or detection, intelligent surveillance, etc. Over the years, many different features have been developed to resolve the problem of no-reference video quality assessment (NR-VQA). In this paper, we propose a novel NR-VQA algorithm that integrates the fusion of temporal statistics of local and global image features with an ensemble learning framework in a single architecture. Namely, the temporal statistics of global features reflect all parts of the video frames, while the temporal statistics of local features reflect the details. Specifically, we apply a broad spectrum of statistics of local and global features to characterize the variety of possible video distortions. In order to study the effectiveness of the method introduced in this paper, we conducted experiments on two large benchmark databases, i.e., KoNViD-1k and LIVE VQC, which contain authentic distortions, and we compared it to 14 other well-known NR-VQA algorithms. The experimental results show that the proposed method is able to achieve greatly improved results on the considered benchmark datasets. Namely, the proposed method exhibits significant progress in performance over other recent NR-VQA approaches.

Evaluation of the Quality and Influence of YouTube as a Source of Information on Robotic Myomectomy

We aimed to evaluate the usefulness of YouTube videos for helping patients become informed about robotic myomectomy. We searched YouTube using "robotic myomectomy" and "Da Vinci myomectomy." Videos were sorted by view count, with the 150 most highly viewed videos being selected. From each video, content type, source, view count, video length, time on YouTube, likes, and dislikes were extracted. A scoring system was used to evaluate video quality. The most prevalent content was recordings of actual surgical procedures of robotic myomectomy, and the most common provider was gynecologic surgeons. Videos directly related to robotic myomectomy were mainly provided by medical groups, had been present on YouTube for a significantly longer time (p = 0.003), and had a higher rate of no responses from viewers (p = 0.014) than videos indirectly related to robotic myomectomy. Videos uploaded by nonmedical groups had more likes, more dislikes, and a higher view ratio (p = 0.029, 0.042, and 0.042, respectively). Scores reflecting video quality did not differ between the two groups. Multiple logistic regression revealed that low-quality videos (less than score 5) were significantly correlated with content indirectly related to robotic myomectomy, poor general quality, fewer views, fewer likes, and no response by viewers. Patients who want to get informed about robotic myomectomy on YouTube should exclude low-quality videos according to such parameters as content, views, and response by viewers. In addition, medical groups should provide videos of good quality for instructing patients about this procedure.

YouTube provides insufficient information on patellofemoral instability.

Videos uploaded to YouTube do not go through a review process, and therefore, videos related to patellofemoral instability may have little educational value. The purpose of this study was to assess the educational quality of YouTube videos regarding patellofemoral instability. A standard search was performed on the YouTube database using the following terms: "unstable kneecap," "patellar instability," "patellofemoral instability," "kneecap dislocation," and "patellar dislocation," and the top 50 videos based on the "relevance" assignment of the YouTube algorithm were included for analysis. The properties, content, and source of each video were recorded. The educational quality of videos was analyzed according to scores obtained using DISCERN, the criteria of Journal of the American Medical Association, Global Quality Score, and Patellofemoral Instability Specific Score, and the quality of the videos was evaluated according to the groupings of these scoring systems. A total of 250 videos were identified, of which 89 were included in the study for analysis. The mean video duration was 11.72 ± 22.03 minutes. The median number of views was 4516.5 (range, 3-6 044 971). The content of the videos was disease-specific in 60%, 20% were related to surgical technique or approach, and 14.1% were exercise videos. Most of the videos were uploaded by physicians (33.7%). The Global Quality Score and DISCERN scores were significantly correlated with video duration. The Patellofemoral Instability Specific Score was significantly correlated with video duration, number of views, view rate, likes, and Video Power Index. According to the DISCERN classification, 69.9% of the videos were very insufficient or insufficient. According to the Patellofemoral Instability Specific Score, 65.2% of videos were evaluated as very low or low. According to the Global Quality Score, 60.7% of videos were rated as poor quality. The quality of YouTube videos about Patellofemoral instability is insufficient. It was found that viewers tend to watch short and low-quality videos.

Analysis of the Operation Mode, Existing Problems, and Solutions of TikTok

In the 21st century, short video platforms began to develop rapidly and gradually replaced text reading. Watching short videos has become a popular entertainment item in people's lives. At the same time, short video platforms began to develop like live e-commerce. This paper discusses the operation mode and target market of TikTok tells people the reasons and methods for the success and rapid development of TikTok and analyzes the future development trend. With the rapid development of these platforms, some problems begin to be exposed. Therefore, this paper will study some problems of the TikTok platform, such as the negative impact of low video quality, and give some relevant solutions, such as increasing audit intensity. In this paper, we find the problems of the TikTok platform and discuss the solutions to the related problems to ensure the normal development of the platform in the future and realize the further expansion of the platform by optimizing and improving platform.

An image enhancement approach for coral reef fish detection in underwater videos

Coral reefs are rich in fisheries and aquatic resources, and the study and monitoring of coral reef ecosystems are of great economic value and practical significance. Due to complex backgrounds and low-quality videos, it is challenging to identify coral reef fish. This study proposed an image enhancement approach for fish detection in complex underwater environments. The method first uses a Siamese network to obtain a saliency map and then multiplies this saliency map by the input image to construct an image enhancement module. Applying this module to the existing mainstream one-stage and two-stage target detection frameworks can significantly improve their detection accuracy. Good detection performance was achieved in a variety of scenarios, such as those with luminosity variations, aquatic plant movements, blurred images, large targets and multiple targets, demonstrating the robustness of the algorithm. The best performance was achieved on the LCF-15 dataset when combining the proposed method with the cascade region-based convolutional neural network (Cascade-RCNN). The average precision at an intersection-over-union (IoU) threshold of 0.5 (AP50) was 0.843, and the F1 score was 0.817, exceeding the best reported results on this dataset. This study provides an automated video analysis tool for marine-related researchers and technical support for downstream applications.

Diverse Pose Lip-Reading Framework

Lip-reading is a technique to understand speech by observing a speaker’s lips movement. It has numerous applications; for example, it is helpful for hearing impaired persons and understanding the speech in noisy environments. Most of the previous works of lips-reading focused on frontal and near frontal face lip-reading and some of them targeted multiple poses in high quality videos. However, their results are not satisfactory on low quality videos containing multiple poses. In this research work, a lip-reading framework is proposed for improving the recognition rate in low quality videos. In this work, a Multiple Pose (MP) dataset of low quality videos containing multiple extreme poses is built. The proposed framework decomposes the input video into frames and enhances them by applying the Contrast Limited Adaptive Histogram Equalization (CLAHE) method. Next, faces are detected from enhanced frames and frontalized the multiple poses using the face frontalization Generative Adversarial Network (FF-GAN). After face frontalization, the mouth region is extracted. The extracted mouth region in the whole video and its respective sentences are then provided to the ResNet during the training process. The proposed framework achieved a sentence prediction accuracy of 90% on a testing dataset containing 100 silent low-quality videos with multiple poses that are better as compared to state-of-the-art methods.

Developing artificial intelligence models for medical student suturing and knot-tying video-based assessment and coaching.

BackgroundEarly introduction and distributed learning have been shown to improve student comfort with basic requisite suturing skills. The need for more frequent and directed feedback, however, remains an enduring concern for both remote and in-person training. A previous in-person curriculum for our second-year medical students transitioning to clerkships was adapted to an at-home video-based assessment model due to the social distancing implications of COVID-19. We aimed to develop an Artificial Intelligence (AI) model to perform video-based assessment.MethodsSecond-year medical students were asked to submit a video of a simple interrupted knot on a penrose drain with instrument tying technique after self-training to proficiency. Proficiency was defined as performing the task under two minutes with no critical errors. All the videos were first manually rated with a pass-fail rating and then subsequently underwent task segmentation. We developed and trained two AI models based on convolutional neural networks to identify errors (instrument holding and knot-tying) and provide automated ratings.ResultsA total of 229 medical student videos were reviewed (150 pass, 79 fail). Of those who failed, the critical error distribution was 15 knot-tying, 47 instrument-holding, and 17 multiple. A total of 216 videos were used to train the models after excluding the low-quality videos. A k-fold cross-validation (k = 10) was used. The accuracy of the instrument holding model was 89% with an F-1 score of 74%. For the knot-tying model, the accuracy was 91% with an F-1 score of 54%.ConclusionsMedical students require assessment and directed feedback to better acquire surgical skill, but this is often time-consuming and inadequately done. AI techniques can instead be employed to perform automated surgical video analysis. Future work will optimize the current model to identify discrete errors in order to supplement video-based rating with specific feedback.

Aerosol Management in Dentistry During the COVID-19 on YouTube

Aim: The novel coronavirus continues to spread around the world despite all efforts. Dental professionals and staff are exposed to high infection risks and have the potential to spread these infections. Web-based portals are widely used in dentistry for obtaining information. YouTube is one of the most popular web platforms provides a wide variety of information, however, concerns are raised regarding the quality of the videos. The purpose of this study is to analyze the content of YouTube videos on aerosol management in dentistry during the COVID-19 pandemic.
 Materials and Methods: YouTube search was performed using the terms “dental aerosol management and COVID-19”, “dental aerosol control and COVID-19” and “dental aerosol reduction and COVID-19” with the default filter set to ‘‘sort by relevance’’. The first 100 videos for each term were viewed and analyzed by 2 independent research. The Global Quality Scale (GQS) and the video information and quality index (VIQI) INDEX were performed to evaluate video quality. Viewers’ interactions and Viewing Rate index were chosen to evaluate the video popularity and also other parameters (number of views, duration, time since upload, and likes and dislikes, country of origin, source of upload) were evaluated. 
 Results: After the exclusion criteria 57 videos were evaluated. 11 of the 57 videos had excellent quality and flow. It was observed that as the duration increased, the quality of the videos in both evaluation indexes (VIQI-GQS) were increased. There was no significant difference in terms of loading sources for VIQI and GQS index.
 Discussion and Conclusions: Besides high-quality videos on aerosol management, there were some low-quality videos that may cause spread of misleading information. Healthcare professionals should play a more active role in the educational videos about infection prevention and aerosol management during pandemic on YouTube.

A Brief Study on Object Detection and Tracking

The detection and tracking of objects in computer vision systems is a critical and challenging subject. Object identification and tracking currently has applications in a wide range of industries, including surveillance, autonomous robot navigation and vehicle navigation because to the increased accessibility of computing power and vast public datasets. When analyzing an image, the object tracking algorithm goal is to divide the region of interest, monitor its motion and location and identify any blobs. Every tracking necessitates an object detection mechanism either in each frame or whenever an object appears for the first time in the video sequence. The availability of high performance processors and low cost-high quality video cameras has enhanced the need of automated object tracking video sequences. Multiple object Identification and tracking in a dynamic environment has recently become focus of computer vision research. In this paper we discuss about the various object detection and tracking algorithms.

Sensitivity-Aware Spatial Quality Adaptation for Live Video Analytics

To address the conflict between the limited network bandwidth and high DNN inference accuracy, live video analytics desires a bandwidth-efficient streaming approach. To this end, more and more works study spatially variable quality streaming where high quality is only used for important regions. The key challenges are to accurately identify the important regions and select the right qualities for them to maximize accuracy. Existing approaches use either cheap analytics models or low-quality videos to locate important regions, and employ heuristic rules to make quality decisions, which struggle to address the above challenges. Our key insight is that the region’s accuracy “sensitivity” obtained by running the expensive DNN model on the high-quality video provides a reliable indication of the region’s importance and allows to allocate the available bandwidth optimally over regions by explicitly maximizing the frame accuracy. This work presents a sensitivity-aware algorithm Orchestra, which incorporates sensitivity into the design of spatial quality adaptation, including video zoning and quality selection. The design of Orchestra entails three main contributions: a feasible way of sensitivity estimation, sensitivity-aware zoning, and deduction-based accuracy estimation. Extensive experiments over realistic videos and network traces show that Orchestra improves accuracy by upto 14.1% with comparable bandwidth usage or reduces bandwidth usage by upto 44.2% while maintaining higher accuracy compared to baselines.

Educational value of spinal injection therapy videos in Korean YouTube for back pain patients.

YouTube, the largest online video platform, has become increasingly popular as a source of health information to patients. The aim of the study was to assess whether Korean patients were well informed about spinal injection from YouTube. Search for the keyword "cheog-chu ju-sa" in Korean language was done, and the quality of the 51 videos with the highest number of views was evaluated independently by two pain management doctors. The averages of global quality scores evaluated by the two doctors were 3.0 and 3.5 and modified DISCERN (mDISCERN) scores were 2.8 and 3.0, respectively. The Kappa statistic between the two doctors' scores was 0.285 and 0.417. The percentage of low-quality videos with a global quality score of 2 or less is 18-36%, which indicated that these videos might provide inaccurate or misleading medical information to the patient. Pain clinic doctors should be wary of medically misleading information available on online platforms, such as YouTube, and strive to create and distribute professional quality educational materials.

Review of the Quality of YouTube Videos Recommending Exercises for the COVID-19 Lockdown.

Background: The world is experiencing a pandemic caused by COVID-19. Insufficient physical activity can increase the risk of illness. Trying to replicate a normal search that any user/patient could do in YouTube, the objective of this study was to evaluate the quality of YouTube videos related to home exercises during lockdown and their adherence to World Health Organization (WHO) recommendations. Methods: A simple search was carried out on YouTube. The first 150 videos were selected. After applying exclusion criteria, 68 videos were analyzed and evaluated. Two statistical analyses based on machine learning techniques were carried out. Videos were classified according to principal component analysis (PCA) models as ‘Relevant’ and ‘Non-Relevant’. Popularity was assessed using the video power index (VPI). Information’s quality and accuracy were gauged using the DISCERN scale and global quality score (GQS). Reliability and credibility of information that can be found on health-related websites was assessed using the Health On the Net Code (HONCode). Exercises were evaluated according to WHO recommendations. Results: DISCERN, HONCode, and GQS scored a mean of 2.29, 58.95, and 2.32, respectively. The PCA calculation allowed videos to auto-classify into high- and low-quality videos. Conclusions: The quality of YouTube videos recommending exercises during lockdown is low and doesn’t reflect WHO recommendations. Effective strategies and tools capable of indicating the quality of this information are needed to filter out erroneous or non-rigorous information that may affect people’s health. These tools should help any user/viewer to distinguish videos of high and low quality.

YouTube as a source of patient education information for elbow ulnar collateral ligament injuries: a quality control content analysis.

BackgroundWhile online orthopedic resources are becoming an increasingly popular avenue for patient education, videos on YouTube are not subject to peer review. The purpose of this cross-sectional study was to evaluate the quality of YouTube videos for patient education in ulnar collateral ligament (UCL) injuries of the elbow.MethodsA search of keywords for UCL injury was conducted through the YouTube search engine. Each video was categorized by source and content. Video quality, reliability, and accuracy were assessed by two independent raters using five metrics: (1) Journal of American Medical Association (JAMA) benchmark criteria (range 0–4) for video reliability; (2) modified DISCERN score (range 1–5) for video reliability; (3) Global Quality Score (GQS; range 1–5) for video quality; (4) ulnar collateral ligament-specific score (UCL-SS; range 0–16), a novel score for comprehensiveness of health information presented; and (5) accuracy score (AS; range 1-3) for accuracy.ResultsVideo content was comprised predominantly of disease-specific information (52%) and surgical technique (33%). The most common video sources were physician (42%) and commercial (23%). The mean JAMA score, modified DISCERN score, GQS, UCL-SS, and AS were 1.8, 2.4, 1.9, 5.3, and 2.7 respectively.ConclusionsOverall, YouTube is not a reliable or high-quality source for patients seeking information regarding UCL injuries, especially with videos uploaded by non-physician sources. The multiplicity of low quality, low reliability, and irrelevant videos can create a cumbersome and even inaccurate learning experience for patients.

The Current State of Vascular Surgery Presence in Bilibili Video Platform of China.

BackgroundWith the development of the Internet, more and more patients search for disease-related information on video platforms during the treatment process, and physicians also look for learning materials through these video platforms. Bilibili is one of the most popular video platforms in China. This study evaluated information on various interesting topics, and related surgical procedures searched through Bilibili.MethodThe Bilibili platform was independently queried for 12 common vascular diseases or related surgical procedures between October and November 2021 by two independent authors using the Baidu search engine. Information about the video and uploader was collected, and descriptive analyses of the overall and first-page results were performed.ResultsA total of 3,998 search results were retrieved by searching 12 vascular-related topics, of which 2,225 actual videos (55.7%) were finally confirmed to be related to medicine. Videos for the public accounted for 84.8% of these 2,225 videos. In addition, 50.5% of the video results were uploaded by vascular surgeons, 12.4% by other specialties, 17.7% by organizations, and 19.4% by other individuals. The total number of videos searched for varicose vein and peripheral vascular diseases was the largest, and the total number of leg amputation videos was the smallest. The largest number of videos for medical professionals was about pulmonary embolism, and the smallest was about leg amputation. On the first pages, 168 results (70.0%) were actually medically relevant, and only 7.7% of the videos were uploaded by vascular surgeons.ConclusionOn the Bilibili platform, videos about vascular diseases are extensive but not comprehensive. The videos uploaded by vascular surgeons are rare, and the results searched are not precise. The online presence of vascular surgeons needs to be improved, which may partially solve the problem of low-quality videos due to the lack of strict management and censorship.