prikaz prve stranice dokumenta Prepoznavanje ljudske aktivnosti u videu primjenom Detectron2 biblioteke i LSTM mreže
Access restricted to authenticated users
undergraduate thesis
Prepoznavanje ljudske aktivnosti u videu primjenom Detectron2 biblioteke i LSTM mreže
2022. urn:nbn:hr:179:864310
Klarić-Kukuz, Matea
University of Split
Faculty of Electrical Engineering, Mechanical Engineering and Naval Architecture
Department of Electronics and Computing

Cite this document

Klarić Kukuz, M. (2022). Prepoznavanje ljudske aktivnosti u videu primjenom Detectron2 biblioteke i LSTM mreže (Undergraduate thesis). Split: University of Split, Faculty of Electrical Engineering, Mechanical Engineering and Naval Architecture. Retrieved from https://urn.nsk.hr/urn:nbn:hr:179:864310

Klarić Kukuz, Matea. "Prepoznavanje ljudske aktivnosti u videu primjenom Detectron2 biblioteke i LSTM mreže." Undergraduate thesis, University of Split, Faculty of Electrical Engineering, Mechanical Engineering and Naval Architecture, 2022. https://urn.nsk.hr/urn:nbn:hr:179:864310

Klarić Kukuz, Matea. "Prepoznavanje ljudske aktivnosti u videu primjenom Detectron2 biblioteke i LSTM mreže." Undergraduate thesis, University of Split, Faculty of Electrical Engineering, Mechanical Engineering and Naval Architecture, 2022. https://urn.nsk.hr/urn:nbn:hr:179:864310

Klarić Kukuz, M. (2022). 'Prepoznavanje ljudske aktivnosti u videu primjenom Detectron2 biblioteke i LSTM mreže', Undergraduate thesis, University of Split, Faculty of Electrical Engineering, Mechanical Engineering and Naval Architecture, accessed 15 November 2024, https://urn.nsk.hr/urn:nbn:hr:179:864310

Klarić Kukuz M. Prepoznavanje ljudske aktivnosti u videu primjenom Detectron2 biblioteke i LSTM mreže [Undergraduate thesis]. Split: University of Split, Faculty of Electrical Engineering, Mechanical Engineering and Naval Architecture; 2022 [cited 2024 November 15] Available at: https://urn.nsk.hr/urn:nbn:hr:179:864310

M. Klarić Kukuz, "Prepoznavanje ljudske aktivnosti u videu primjenom Detectron2 biblioteke i LSTM mreže", Undergraduate thesis, University of Split, Faculty of Electrical Engineering, Mechanical Engineering and Naval Architecture, Split, 2022. Available at: https://urn.nsk.hr/urn:nbn:hr:179:864310

Please login to the repository to save this object to your list.
Metadata
TitlePrepoznavanje ljudske aktivnosti u videu primjenom Detectron2 biblioteke i LSTM mreže
Title (english)Human action recognition using Detectron2 and LSTM network
AuthorMatea Klarić-Kukuz
MentorMatko Šarić https://orcid.org/0000-0001-9752-9653 (mentor)
GranterUniversity of Split
Faculty of Electrical Engineering, Mechanical Engineering and Naval Architecture
(Department of Electronics and Computing)
Split
Defense date and country2022-09-22, Croatia
Scientific / art field,
discipline and subdiscipline		TECHNICAL SCIENCES
Computing
Abstract
U ovom radu za prepoznavanje ljudske akitvnosti koristimo biblioteku Detectron2 i LSTM mreže.Detectron2 je biblioteka za otkrivanje ključnih točak na ljudskom tijelu.Zatim se koristi trenirani LSTM model za klasifikaciju radnji.Cjevovod zaključivanja sastoji se od modela Detectron2 i prilagođenog LSTM modela.Aplikacija funkcionira tako da prihvaća video ulaz, ponavlja okvire i zatim koristi Detectron2 za otkrivanje ključnih točaka na svakom okviru.Rezultati ključnih točaka se dodaju u međuspremnik i sadržaj međuspremnika se šalje treniranom LSTM modelu za identifikaciju akcije. Web aplikacija temeljena na Flasku ima korisničko sučelje za prihvaćanje video unosa od korisnika. Radnje koje otkrio cjevovod zaključivanja označene su na videozapisu i prikazane kao rezultat. Što se tiče koda imamo primjere kodiranje ključnih komponenti i prikaz onog što bi aplikacija trebala raditi.Aplikaciju pokrećemo u Google Colabu i kada pokrenemo imamo priliku analizirati pokrete nekog svog video zapisa ili imamo primjer iz aplikacije.
Abstract (english)
In these thesis to recognize human activity, we use the Detectron2 library and LSTM. Detectron2 is a library for detecting key points on the human body. Then a trained LSTM model is used to classify actions. The inference pipeline consists of the Detectron2 model and a custom LSTM model. The application works by accepting a video input , iterates through the frames and then uses Detectron2 to detect keypoints on each frame. The keypoint results are added to a buffer and the contents of the buffer are sent to the trained LSTM model for action identification. The Flask-based web application has a user interface for accepting video input from users. Actions detected by the inference pipeline are highlighted on the video and displayed as output. As for the code, we have examples of the coding of key components and a display of what the application should do. We launch the application in Google Colab and when we launch it, we have the opportunity to analyze the movements of one of our videos or we have an example from the application.
Keywords
Keywords (english)
Languagecroatian
URN:NBNurn:nbn:hr:179:864310
Study programmeTitle: Computing
Study programme type: university
Study level: undergraduate
Academic / professional title: sveučilišni/a prvostupnik/ prvostupnica (baccalaureus/baccalaurea) inženjer/inženjerka računarstva (sveučilišni/a prvostupnik/ prvostupnica (baccalaureus/baccalaurea) inženjer/inženjerka računarstva)
Type of resourceText
File originBorn digital
Access conditionsAccess restricted to authenticated users
Terms of useLicence In copyright icon
Created on2023-05-22 11:32:45