Training Model Research Articles

Decoding Continuous Tracking Eye Movements from Cortical Spiking Activity.

Eye movements are the primary way primates interact with the world. Understanding how the brain controls the eyes is therefore crucial for improving human health and designing visual rehabilitation devices. However, brain activity is challenging to decipher. Here, we leveraged machine learning algorithms to reconstruct tracking eye movements from high-resolution neuronal recordings. We found that continuous eye position could be decoded with high accuracy using spiking data from only a few dozen cortical neurons. We tested eight decoders and found that neural network models yielded the highest decoding accuracy. Simpler models performed well above chance with a substantial reduction in training time. We measured the impact of data quantity (e.g. number of neurons) and data format (e.g. bin width) on training time, inference time, and generalizability. Training models with more input data improved performance, as expected, but the format of the behavioral output was critical for emphasizing or omitting specific oculomotor events. Our results provide the first demonstration, to our knowledge, of continuously decoded eye movements across a large field of view. Our comprehensive investigation of predictive power and computational efficiency for common decoder architectures provides a much-needed foundation for future work on real-time gaze-tracking devices.

Factors associated with the implementation of dialectical behavior therapy by Spanish speaking mental health professionals who treat suicide risk.

The lack of training for professionals on how to manage suicide risk is an important barrier to effective intervention. Dialectical Behavior Therapy Intensive Training™ (DBT-IT) includes specific training for suicide and has shown promising results to enhance implementation of DBT. To our knowledge, no published studies have evaluated the effect of DBT-IT on therapists' attitudes towards treating suicide risk and among Spanish-speaking mental health professionals. The main aim of this study was to evaluate the effect of DBT-IT on therapists' attitudes regarding treating suicide risk and its relationship with the implementation of DBT before and after receiving DBT-IT. A total of 242 mental health workers (76.4% women, mean age 35.38, SD = 9.17; 77.7% from Latin America; 22.3% from Spain) who had received a DBT-IT participated in the study. Self-efficacy (Efficacy in Assessing and Managing Suicide Risk Scale) and concerns (Concerns about Treating Suicidal Clients Scale) in treating suicide, perceived burnout (Copenhagen Burnout Inventory), confidence in applying DBT (Behavioral Anticipation and Confidence Questionnaire), barriers to implementation (Barriers to Implementation Inventory), implementation of DBT and reach were measured via online survey at parts 1 (pre-training) and 2 (post-training, after 9 months of implementation) of the DBT-IT. Differences between pre-training and post-training (n = 61) indicated statistically significant improvements in self-efficacy, concerns about the lack of training and competence in treating suicide, and confidence in applying DBT. Statistically significant increases in the rates of DBT treatment modes implementation (except for individual therapy) and mindfulness practice, as well as the number of team members and consultation team hours, were also found. Findings also indicated statistically significant positive correlations between burnout and concerns about treating suicidal clients, as well as with structural and administrative implementation barriers, and between self-efficacy in managing suicide, confidence in applying DBT and implementation of DBT treatment modes. Participants with more implementation barriers reported lower rates of consultation team meetings and phone coaching implementation. DBT-IT could be an adequate training model to increase self-efficacy and confidence in treating suicide risk as well as to facilitate implementation of DBT treatment modes by Spanish-speaking mental health professionals. Not applicable.

Evaluating the effectiveness of a novel training model in dental preclinical regenerative endodontic education.

Regenerative endodontic procedures (REPs) are tissue engineering-based treatments aimed at regenerating the pulp-dentin complex. However, these procedures are not widely practiced or well-known among dentists due to a lack of training. This study aims to develop a practical hands-on training model for REPs to enhance students' theoretical knowledge and practical skills. A cost-effective, easy-to-operate, and reusable training model for preclinical training of REPs was developed. A total of 125 undergraduates were divided into an experimental group (REPs model) and a control group (traditional teaching). The students' theoretical knowledge and operational skills were assessed through examinations and practical assessments. Student perceptions were assessed by questionnaires. The experimental group achieved higher scores in theoretical knowledge and operational skills compared to the control group (p < .001). The process evaluation of the experimental group demonstrated a high level of proficiency in executing root canal disinfection steps, with pass rates ranging from 93.5% to 95.2%. Students in both groups perceived the model as valuable for improving critical thinking, stimulating interest, and enhancing their confidence and proficiency. The developed REPs training model is a practical and effective tool for teaching and training regenerative endodontics. It bridges the gap between preclinical and clinical education, enhancing students' confidence and clinical competence.

Analysis of impact of limb segment length variations during reinforcement learning in four-legged robot.

Crawling robots are becoming increasingly prevalent in both industrial and private applications. Despite their many advantages over other robot types, they have complex movement mechanics. Artificial intelligence can simplify this by reinforcement learning. This process requires configuring the training environment and defining input parameters, including a robot model for movement training. To translate the virtual results into real-world scenarios, a 3D model with appropriate mechanical parameters must be developed.These parameters can vary significantly between multiple mechanical configurations, which will further impact the reinforcement learning process of such a robot. For this reason, it was decided to test which limb configurations would work best in this process. Initially, various kinematic types of walking robots were analysed, drawing on the anatomy of mammals, reptiles, and insects for the biological model. The reptilian model was chosen for its balance of stability, dynamics, and energy efficiency. The article reviews the preparation of robot models and the configuration of the Unity3D development environment using the ML-Agents toolkit. The experiment examined how different limb lengths affect training, resulting in movement algorithms for various quadruped robot configurations using artificial neural networks. Based on the numerical results, the best configuration was the default, with the same length of the tibia as the thigh, achieving a reward function value of 883.9 and an episode length of 245.5. Taking into account the same criteria, the least efficient configuration was definitely the one characterised by the shortest thigh and the longest tibia among those considered. In its case, the reward function reached a value of only 526.2 with an episode lasting 999.0, which means that it never achieved the intended goal.

Development and validation of a cost-effective DIY simulation model for McDonald cerclage training.

The prevention of preterm birth is a challenging task for obstetricians. Cervical cerclage, used as both a primary and secondary prevention method for spontaneous preterm birth, is a crucial surgical intervention. It is essential that obstetricians can learn this procedure in a simulated environment before performing the stitches on high-risk patients. This study aimed to develop a simulator based on 3D printing and evaluate its validity for clinical training. The objectives of this study were (1) to design and construct a cost-effective simulator for McDonald cerclage with two different cervix models-a closed cervix and a cervix with bulging membranes-using common material from a DIY store and 3D printing technology and (2) to validate its effectiveness through feedback from learners and experts in cervical cerclage. The self-made simulator was evaluated by obstetricians using a questionnaire with Likert scale. Obstetricians and gynecologists assessed the simulator and found it useful for learning and practicing cervical cerclage. The simulator was deemed valuable for skill training. Cervical cerclage is a complex procedure that should be mastered through simulation rather than initial practice on real patients. Our simulator is a cost-effective model suitable for various clinical settings. It has been validated by obstetricians for both preventive and therapeutic cerclage, demonstrating its efficacy for training in cerclage techniques. Future research should focus on less skilled obstetricians and gynecologists and investigate how repeated use of the simulator can enhance their performance in cerclage stitching.

Accuracy of O-RADS System in Differentiating Between Benign and Malignant Adnexal Masses Assessed via External Validation by Inexperienced Gynecologists

Objective: To evaluate the accuracy of the O-RADS system in differentiating between benign and malignant adnexal masses, as assessed by inexperienced gynecologists. Methods: Ten gynecologic residents attended a 20 h training course on the O-RADS system conducted by experienced examiners. Following the training, the residents performed ultrasound examinations on patients admitted with adnexal masses under supervision, recording the data in a database that included videos and still images. The senior author later accessed this ultrasound database and presented the cases offline to ten residents for O-RADS rating, with the raters being blinded to the final diagnosis. The efficacy of the O-RADS system by the residents and inter-observer variability were assessed. Results: A total of 201 adnexal masses meeting the inclusion criteria were evaluated, consisting of 136 (67.7%) benign masses and 65 (32.3%) malignant masses. The diagnostic performance of the O-RADS system showed a sensitivity of 90.8% (95% CI: 82.2–96.2%) and a specificity of 86.8% (95% CI: 80.4–91.8%). Inter-observer variability in scoring was analyzed using multi-rater Fleiss Kappa analysis, yielding Kappa indices of 0.642 (95% CI: 0.641–0.643). The false positive rate was primarily due to the misclassification of solid components in classic benign masses as O-RADS-4 or O-RADS-5. Conclusions: The O-RADS system demonstrates high diagnostic performance in distinguishing benign from malignant adnexal masses, even when used by inexperienced examiners. However, the false positive rate remains relatively high, mainly due to the over-interpretation of solid-appearing components in classic benign lesions. Despite this, inter-observer variability among non-expert raters was substantial. Incorporating O-RADS system training into residency programs is beneficial for inexperienced practitioners. This study could be an educational model for gynecologic residency training for other systems of sonographic features.

Research on Key Technologies of Image Steganography Based on Simultaneous Deception of Vision and Deep Learning Models

As machine learning continues to evolve, traditional image steganography techniques find themselves increasingly unable to meet the dual challenge of deceiving both the human visual system and machine learning models. In response, this paper introduces the Visually Robust Image Steganography (VRIS) model, specifically tailored for this dual deception task. The VRIS model elevates visual security through the use of specialized feature extraction and processing methodologies. It meticulously conducts feature-level fusion between secret images and cover images, ensuring a high level of visual similarity between them. To effectively mislead machine learning models, the VRIS model incorporates a sophisticated strategy utilizing random noise factors and discriminators. This involves adding controlled amounts of random Gaussian noise to the encrypted image, thereby enhancing the difficulty for machine learning frameworks to recognize it. Furthermore, the discriminator is trained to discern between the noise-infused encrypted image and the original cover image. Through adversarial training, the discriminator and VRIS model refine each other, successfully deceiving the machine learning systems. Additionally, the VRIS model presents an innovative method for extracting and reconstructing secret images. This approach safeguards secret information from unauthorized access while enabling legitimate users to non-destructively extract and reconstruct images by leveraging multi-scale features from the encrypted image, combined with advanced feature fusion and reconstruction techniques. Experimental results validate the effectiveness of the VRIS model, achieving high PSNR and SSIM scores on the LFW dataset and demonstrating significant deception capabilities against the ResNet50 model on the Mini-ImageNet dataset, with an impressive misclassification rate of 99.24%.

Multi-Model Style-Aware Diffusion Learning for Semantic Image Synthesis

Semantic image synthesis aims to generate images from given semantic layouts, which is a challenging task that requires training models to capture the relationship between layouts and images. Previous works are usually based on Generative Adversarial Networks (GAN) or autoregressive (AR) models. However, the GAN model's training process is unstable, and the AR model’s performance is seriously affected by the independent image encoder and the unidirectional generation bias. Due to the above limitations, these methods tend to synthesize unrealistic, poorly aligned images and only consider single-style image generation. In this paper, we propose a Multi-model Style-aware Diffusion Learning (MSDL) framework for semantic image synthesis, including a training module and a sampling module. In the training module, a layout-to-image model is introduced to transfer the learned knowledge from a model pretrained with massive weak correlated text-image pairs data, making the training process more efficient. In the sampling module, we designed a map-guidance technique and creatively designed a multi-model style-guidance strategy for creating images in multiple styles, e.g., oil painting, Disney Cartoon, and pixel style. We evaluate our method on Cityscapes, ADE20K, and COCO-Stuff, making visual comparisons and computing with multiple metrics such as FID, LPIPS, etc. Experimental results demonstrate that our model is highly competitive, especially in terms of fidelity and diversity.

Content Validity for a 3D-Printed Model for Simulation-Based Training of Basic Epistaxis Management

Introduction Hands-on training of basic epistaxis management is often minimal during pregraduate medical training. The objective was to develop and gather structured content validity evidence for a 3D-printed model for simulation-based training and assessment of technical skills in basic epistaxis management. Methods A simulator for epistaxis management training was built based on publicly available 3D-print files with addition of tubing to mimic anterior bleeding. Ten otorhinolaryngologists evaluated the model after performing nasal cavity inspection and insertion and insufflation of a nasal tamponade device. Content validity evidence was collected following Messick's framework. Results Two content experts contributed to define features in the iterative building process. In the structured evaluation of the model, experienced clinicians found the appearance of the outer nose and the resistance when insufflating a nasal tamponade device into the nasal cavity satisfying (mean score of 3.9 and 4.7 out of 5, respectively), whereas the nasal cavity was found to be a bit too spacious compared with typical real-life conditions. Conclusions A 3D-printed simulator for simulation-based training of basic epistaxis management was successfully built. Content validity was gathered, and overall, content experts found the model to adequately represent the technical skills challenges for training novices such as medical students and junior doctors.

Assessing the Use of a Training Model Using Immersive Virtual Reality for Sacrospinous Fixation Surgery: A Pilot Study.

Sacrospinous fixation (SSF) is a surgical complex procedure for uterovaginal prolapse that may result in pudendal plexus and vascular damage. The objective was to evaluate a training model using virtual reality to perform SSF. We conducted a prospective single-center pilot study in the university's academic innovation department in November 2023. We recruited 20 participants, 10 gynecology resident surgeons (group 1) and 10 experienced gynecology surgeons (group 2). The mean age was 26.6 ± 0.6years in group 1 and 39.7 ± 3.1years in group 2. The satisfaction rate reached 90% in both groups. A scale from 1 to 4 (1 = not-at-all, 4 = completely) was used to assess participants' perceptions. The realism of the simulation was rated at 3.3 ± 0.1 in group 1 and 3.5 ± 0.2 in group 2 (p = 0.41), the haptic feedback was rated at 2.0 ± 0.3 and 1.8 ± 0.3 (p = 0.66), the mean exercise time was 6.4 ± 1.3min and 4.8 ± 0.7min (p = 0.58), and the mean success rate was 70.8 ± 3.2% and 83.5 ± 6.8% (p = 0.14). In group 2, left and right instruments were placed at 2.6 ± 0.8mm and 2.9 ± 0.7mm from the target respectively, and in group 1 at 4.8 ± 0.7mm and 3.8 ± 0.5mm respectively. Participants reported greater comfort in performing sacrospinous fixation after simulation training (3.0 ± 0.2 versus 2.3 ± 0.3, p = 0.17). Our virtual reality simulation model is feasible and accurate for SSF training for both residents and experienced surgeons with a high level of satisfaction. Virtual reality simulation improves technical skills and promotes surgical comfort.

A neurosurgical nursing training system based on 3D printing models: practice and exploration of precision medicine.

In recent years, the concept of precision medicine has gradually permeated the field of nursing, presenting new challenges for neurosurgical nursing staff. Traditional teaching methods often struggle to adequately convey complex anatomical knowledge and nursing skills, thereby affecting the learning outcomes and clinical decision-making abilities of nurses. As an innovative educational tool, three-dimensional (3D) printing technology has demonstrated significant potential in medical education in recent years. Therefore, this project aims to explore the advantages and feasibility of using 3D printed models as a novel teaching tool in neurosurgical nursing. The study involved 80 nursing personnel who participated in the neurosurgical rotation training at Shaoxing People's Hospital from 2022 to 2023. The control group (40 individuals) used only traditional teaching tools such as textbooks, multimedia courseware, and laboratory equipment for theoretical and practical training based on the curriculum. The experimental group (40 individuals) incorporated a 3D printing model database as an innovative teaching tool on the basis of traditional teaching methods, and used patient-specific 3D printed models for precise nursing training. After the training, both groups were assessed for theoretical and practical abilities, professional values, job satisfaction, and the prognosis of intracranial aneurysm patients they cared for. The experimental group demonstrated superior scores in theoretical and practical abilities, professional values, and job satisfaction compared to the control group. During the training period, patients with intracranial aneurysms receiving precision nursing care from the experimental group showed better quality of life and neurological functional outcomes compared to the control group. The nursing training system based on 3D printing models can enhance the learning efficiency and nursing work satisfaction of nursing personnel. It may have a positive impact on the professional values of nursing personnel, thereby laying a solid foundation for providing high-quality, precise nursing services in the field of neurosurgery in the future.

From Theory to Practice: Virtual Reality Technology Facilitates Efficient Transformation of Preschool Teacher Education

This paper explores the application of Virtual Reality (VR) technology in the efficient transformation of preschool teacher education, from theory to practice. The research background highlights the importance of preschool education, the crucial role of teacher training in enhancing its quality, and the potential of VR technology in education. The study aims to investigate how VR technology can facilitate the efficient transformation of preschool teacher education, analyze its specific applications and effects, and propose strategies and recommendations for teacher training based on VR technology. The literature review summarizes current research on VR technology in education, progress in preschool teacher training models, and preliminary explorations of VR in teacher training. The theoretical foundation is based on learning theories such as constructivism and situational learning, technology acceptance and use theory, and theories of educational innovation and technology integration. An analysis of the current situation in preschool teacher training reveals limitations in traditional models, such as the disconnection between theory and practice, the lack of immersive learning experiences, and inadequate personalized training. The paper then discusses the advantages of VR technology, including immersive learning, flexible teaching scenarios, and personalized learning paths. The conclusion summarizes the value of VR technology in preschool teacher training, successful experiences, and existing problems, and provides future directions for the deep integration of technology and education, continuous innovation in teacher training models, and implications for educational policy and practice.

Development and Validation of a Simulation Model for Ureteral Stent Placement.

Currently, there are no validated training models for cystoscopy with ureteral stent placement. The objectives of this study were to develop and validate a novel endoscopic simulation model for training in ureteral stent placement. A low-cost, low-fidelity training model was developed to simulate ureteral stent placement. Recruited participants were divided into 3 groups: novices (postgraduate year 3/4 gynecology residents), advanced learners (urogynecology and reconstructive pelvic surgery fellows), and experts (urology residents, urogynecology faculty, and urology faculty). Construct validity was measured using de-identified video-recorded performances on the model, which were evaluated by 2 expert reviewers using validated scales (Global Operative Assessment of Laparoscopic Skills [GOALS], Global Rating Scale [GRS]) and procedure-specific metrics. The model was created using a hollow Styrofoam sphere, plastic tubing from a retropubic sling, and a silicone pacifier. Thirty-six surgeons were assessed performing the procedure using the model with cystoscopic equipment. The experts (n = 12) performed significantly better than the advanced learners (n = 17) and novices (n = 7) in total scores (max 75, median [IQR]: 75 [75-75], 61 [56.5-68.5], 45 [43-46], respectively; P < 0.001) and within each individual scale domain. Increasing experience with ureteral stent placement had a significant correlation (P < 0.001) with better performance on the model. A minimum total passing score of 63 was established. On post simulation assessment, most participants "agreed" or "strongly agreed" that the model closely approximates the feel of ureteral stent placement. This ureteral stenting simulation model is easy to construct, affordable, and reproducible. The model is valid and reliable for practicing the procedure in preparation for live surgery.

Evaluation of training models for intraventricular neuroendoscopy.

Structured surgical education has become increasingly important in recent years. Intraventricular neuroendoscopic procedures have been widely established. However, training surgical skills with these techniques is crucial for young residents due to the potential harm to adjacent tissue. Therefore, we evaluated two different training models.Participants in two different international workshops were trained on a prefixed cadaver model and on a living murine intraabdominal model. Crucial neuroendoscopic techniques such as membrane perforation and tissue biopsy were performed. A blinded questionnaire evaluated both models.Sixty-three participants were trained on the animal model. Forty of these were trained on the cadaver model. The training effect was evaluated almost equally, with 8.5/10 for the animal model and 8.9/10 for the cadaver model. The tissue properties were rated higher regarding realism in the animal model, whereas the anatomic realism was rated higher in the cadaver model.The animal model is a valid alternative to cadaver models for teaching endoscopic neurosurgical skills. This model benefits from the simulation of real surgical tissue properties, including bleeding. The low costs and availability of this technique make it more ubiquitous and can help train further generations of neurosurgeons.

Knowledge-based Cyber Physical Security at Smart Home: A Review

Smart-home systems represent the future of modern building infrastructure as they integrate numerous devices and applications to improve the overall quality of life. These systems establish connectivity among smart devices, leveraging network technologies and algorithmic controls to monitor and manage physical environments. However, ensuring robust security in smart homes, along with securing smart devices, presents a formidable challenge. A substantial number of security solutions for smart homes rely on data-driven approaches (e.g., machine/deep learning) to identify and mitigate potential threats. These approaches involve training models on extensive datasets, which distinguishes them from knowledge-driven methods. In this review, we examine the role of knowledge within smart homes, focusing on understanding and reasoning regarding various events and their utility toward securing smart homes. We propose a taxonomy to characterize the categorization of decision-making approaches. By specifying the most common vulnerabilities, attacks, and threats, we can analyze and assess the countermeasures against them. We also examine how smart homes have been evaluated in the reviewed literature. Furthermore, we explore the challenges inherent in smart homes and investigate existing solutions that aim at overcoming these limitations. Finally, we examine the key gaps in smart-home-security research and define future research directions for knowledge-driven schemes.

FORMAÇÃO CONTINUADA DE PROFESSORES NO MUNICÍPIO DE VENDA NOVA DO IMIGRANTE: UMA CARACTERIZAÇÃO

This work is an excerpt from the master's dissertation entitled Collaborative continuing education: an experience with physical education teachers in Venda Nova do Imigrante. This text does not intend to exhaust the information about education in the municipality, but only to briefly characterize the history of education in the municipality. The study points out the development process of the region since the arrival of the first immigrants, through the establishment of its first school, and continuing to education in the present day, pointing out the training model adopted by the municipality in recent years according to official documents.

The Pros and Cons of the Piano Accelerated Training Model for Art Students

In the Guangdong Province Music College Entrance Examination, piano, as the main and secondary examination subject, plays a pivotal role for candidates. At present, piano candidates are mainly divided into two categories: one is candidates who have learned piano since childhood and have accumulated a deep basic skill and theoretical foundation; the other is candidates who start to practice piano for personal interest or other reasons when they are about to take the college entrance examination. Piano crash course has both advantages and disadvantages for art exam students. The advantage is that crash course learning allows candidates to master certain piano skills and knowledge in a short period of time, and improve their understanding and test-taking ability. The disadvantage is that due to the lack of long-term basic practice and a large number of works training accumulation, candidates may lack depth and breadth in piano performance, and their understanding and mastery of works may not be deep enough. Therefore, among art exam students who choose piano crash course learning, it is necessary to focus on the training and accumulation of works while mastering basic knowledge and skills, as well as deeply understand the connotation of the works. Only in this way can we achieve excellent results in the art college entrance examination and continue to develop on the road of music in the future.

Threshold Filtering for Detecting Label Inference Attacks in Vertical Federated Learning

Federated learning, as an emerging machine-learning method, has received widespread attention because it allows users to train locally during the training process and uses relevant cryptographic knowledge to safeguard the privacy of data during model aggregation. However, existing federated learning is also susceptible to privacy breaches, e.g., label inference attacks against vertical federated learning scenarios, where an adversary is able to reason about the labels of other participants based on the trained model, leading to serious privacy breaches. In this paper, we design a detection method for label inference attacks in vertical federated learning scenarios, which is able to detect the attacks based on the principles of the attacks. We design a threshold-filtering detection method based on the principle of attack to determine that the model is under attack when the threshold value is greater than a set parameter. Furthermore, we have created six threat model classifications based on different a priori conditions of the adversary to comprehensively analyze the adversary’s attacks. In addition to the detection method of attacks, the extent of attacks on the model and the effectiveness of the defense can also be evaluated. The evaluation module will experimentally measure the changes in the relevant metrics such as the accuracy of the attack, the F1 score, and the change in the accuracy after the defense method. For example, detection in the full connected neural network model assesses the attack and defense effectiveness of the model with an attack accuracy of 86.72% in the breast cancer Wisconsin dataset and an F1 score of 0.743, which is reduced to 36.36% after dispersed training. This ensures that users have an overall grasp of the extent to which the training model is under attack before deploying the model.

Participatory research at scale: A comparative analysis of four approaches to large-scale agricultural technology testing with farmers

Tailoring agricultural technology options to the diverse conditions of smallholder farmers requires innovative approaches for testing these technologies with farmers across varied contexts, while incorporating their feedback into learning and decision-making processes. This study compares four such approaches: the Farmer Field School on Participatory Plant Breeding (FFS-PPB), Farmer Research Network (FRN), Crowdsourcing–Triadic comparisons of technologies (Tricot), and adapted Mother–Baby Trial (MBT) as implemented by four concrete projects. The objectives are to provide detailed descriptions of these approaches and their project-specific implementations, identify and analyze common aspects and differences, and derive insights to guide future farmer-inclusive projects aiming at contextual scaling of agricultural technologies. A literature review, key informant interviews, and a systematic content analysis were conducted for the analysis. Common features include cascade training models, simple farmer-managed experiments, and the use of digital tools for data collection. Major differences lie in the extent of farmer–researcher collaboration and decision-making, as well as how technology option-by-context interactions are addressed. The FRN, FFS-PPB, and adapted MBT approaches involve farmers in decision-making throughout most stages of research, including co-learningcycles that adapt the research design and technology options to farmers’ needs. Although these approaches require more training and expertise, they increase the likelihood of achieving relevant results that farmers can implement in practice. In contrast, more standardized approaches like the Crowdsourcing–Tricot streamline the implementation, data management and analysis of large-scale trials, but have limitations in capturing the underlying reasons for farmers’ preferences. Among the studied approaches, the FRN as implemented by the Women's Fields project in Niger is particularly effective in identifying which options best suit specific farming contexts.

An improved face attributes editing method based on DDIM

The main advantage of DDIM is that it guarantees the quality of the generated images while increasing the efficiency of the generation by modifying the sampling strategy in the diffusion process. DiffusionRig, which addresses the problem of maintaining identity consistency by learning the person-specific facial prior in a tiny personalized dataset, is a successful representation of the DDIM strategy. Based on DiffusionRig, in this article, we propose an improved face attributes editing method based on DDIM to improve naturalness and accuracy of editing results in complex face attribute editing tasks and the generalization ability. Our method combines DDIM and DECA together and use two-stage training strategy. To reduce DiffusionRig’s limitations in handling face attribute editing tasks that require nonlinear understanding and fine-tuning, our method also introduces a channel attention mechanism and a depth-separable convolution technique in the training model. In the first stage we trained our model on the open FFHQ dataset, which consists of 30,000 high-resolution face image. In the second stage, the model was refined by using a tiny personalized dataset. Face attribute editing experiments,comparative experiment with DiffusionRig, and a series of ablation experiments have been performed. Then, the validity of our improved method is verified by qualitative and quantitative analysis.