Aban, Steven Claude N.; Baldonado, Greian Kif A.; Muro, Cyron Paul T.; Rodriguez, Ralph Emmanuel D.; Tecson, Christian James G. and Tuazon, Morthon Jhayzie.  4  0   <br/><br/>
Billi: Wearable, AI assisted bill identifier for the blind /        6 
       6     
Aban, Steven Claude N.; Baldonado, Greian Kif A.; Muro, Cyron Paul T.; Rodriguez, Ralph Emmanuel D.; Tecson, Christian James G. and Tuazon, Morthon Jhayzie. 
 -   
 -     
 - 82 pp.     28 cm.     
 -   
 -   
 - 
 - 
 -         .
 -     .
 -          0   .
 -         .
 -  0       .
<br/><br/>   Undergraduate Thesis: (Bachelor of Science in Computer Engineering) - Pamantasan ng Lungsod ng Maynila, 2024. 
<br/><br/>
<br/><br/>      
<br/><br/>   5 
<br/><br/>      
<br/><br/> ABSTRACT:  STATEMENT OF THE PROBLEM:  	Visual impairement makes identifying paper bills difficult for individual's, lacking a practical solution. Current methods are unreliable and compromise privacy. An innovative wearable AI bill identifier is needed to empower the blind with financial independence.  RESEARCH METHODOLOGY:  	This study aims to aid visually impaired individuals in managing paper money using Billi, Employing a mixed-methods approach, the study begins with qualitative user-centered design and participatory techniques. Subsequently, a controlled experimental design involves blind participants in simulated transactions. Quantitative data will be complemented by user surveys, interviews, and usability testing. The comprehensive evaluation ensures both user perspectives and quantitative measures contribute to Billi's development and integration into visually impaired individual's daily lives.  SUMMARY OF FINDINGS:  	In a early training, the model had high loss values, expected as it learned. By step 4,500, loss decreased significantly, showing improved understanding. This trend continued, with further reduction at steps 10,000 and 15,000, indicating progress. Regularization loss decreased consistently from step 100 to 20,100, reducing overfitting and improving generalization. Learning rate initially cautious, increased gradually, but fluctuated at step 15,100, highlighting potential instability. At step 20,000, a significant decrease in learning rate suggested possible overshooting. The detection model achieved an impressive total accuracy rate of 91.3% in identifying different Philippine backnotes, with consistent high accuracy rates across all denominations, demonstrating reliability and efficiency.  CONCLUSION:  	The Billi device offers a significant solution for visually impaired individuals in financial transactions. Our study shows iys accuracy, particularly with lower denominations like Php20, Php100, and Php200 notes, reaching perfect scores. While slightly lower for larger denominations, ranging from 80% to 88%, the device still performs well overall. Users rate it highly for efficacy, confidence, and user -friendliness, indicating its potential to foster inclusivity and independence. Additionally, our evaluation highlights the device's precision in identifying Philippine banknotes, suggesting efficiency improvements in financial processing and anti-counterfeiting efforts. Overall, the study underscores the transformative impact of the Billi device in enhancing financial autonomy for visually impaired individuals, emphasizing the need for continued development and implementation to ensure accessibility and equality.  RECOMMENDATION:  	Enhancing device portability by using smaller materials or hardware components without sacrificing performance. Integrating audio feedback with sufficient volume for blind users, improving convenience. Ensuring the device can detect and identify the new Php1000 bill variation, considering its unique security features. Optimizing hardware selection to balance cost and performance, avoiding limitations in functionality and compatibility with packages like TensorFlor and OpenCV. Training the model in various light conditions to improve bill identification accuracy.       
<br/><br/>      
<br/><br/><br/> <br/> <br/> <br/> <br/> <br/> <br/> 
<br/><br/><br/>5     
<br/><br/><br/> <br/>   
<br/><br/><label>ISBN: </label>  
<br/><br/><label>ISSN: </label>2 
 =   
 =   
<br/><br/><label>Standard No.: </label>        
<br/><br/><label>Publisher. No.: </label>  
<br/><br/><label>Source: </label>        
<br/><br/><label>LCCN: </label>
<br/><br/><label>Nat. Bib. No.: </label> 2 
<br/><br/><label>Nat. Bib. Agency Control No.: </label>    
<br/><br/><label>Subjects--Personal Names: </label><br/>     2     --0------
<br/><br/><label>Subjects--Corporate Names: </label><br/>     6 --0-- 2 --------
<br/><br/><label>Subjects--Meeting Names: </label><br/><br/> 0    2 --
<br/><br/><label>Subjects--Uniform Titles: </label><br/>  --20------
<br/><br/><label>Subjects--Chronological Terms: </label><br/> 
<br/><br/><label>Subjects--Topical Terms: </label><br/>    --------20--
<br/><br/><label>Subjects--Geographic Terms: </label><br/>--------20--
<br/><br/><label>Index Terms--Genre/Form: </label><br/>----2
<br/><br/><label>LC Class. No.: </label> / 2 
<br/><br/><label>Dewey Class. No.: </label> / 2 
<br/><br/><label>National Agricultural Library Call No.: </label> /  
<br/><br/><label>National Library of Medicine Call No.: </label> /  
<br/><br/><label>Gov. Doc. Class. No.: </label>