Browsing by Author "Kumar, A"

Filter results by typing the first few letters
Now showing 1 - 4 of 4
  • Thumbnail Image
    PublicationEmbargo
    Emergency Communication Application for Speech and Hearing-Impaired Citizens
    (IEEE, 2020-12-15) Dewasurendra, D; Kumar, A; Perera, I; Jayasena, D; Thelijjagoda, S
    Citizens worldwide with mutism and hearing loss, converse differently than the rest of the population. Preferring a gestural mode of communication, they use sign languages which are established natural languages with an extensive vocabulary and syntax. Nevertheless, despite having equal status to spoken languages, society prominently caters to the majority of citizens who are capable of speaking and processing spoken languages. As a result, sign language speakers face difficulties with social aggregation. The Sri Lankan citizens with mutism and hearing loss obtain improving benefits with the advancement of technology; however, accessing emergency services remains a challenge as they can only be accessed verbally using a phone call. Project `Wadhan' enables citizens with mutism and hearing loss to request for emergency services in real-time without depending on an intermediary. This cross-platform mobile application will convert Sri Lankan Sign Language (SSL) statements to voice and vice-versa, which allows a seamless conversation between the SSL speaker and the emergency operator. The application includes SSL Gesture Recognition, Sinhala Text-to-Speech (TTS), Sinhala Speech Recognition, and SSL Animation Generation, thus resulting in an emergency communication application that can be used by SSL users in dire situations.
  • Thumbnail Image
    PublicationOpen Access
    Fabrication of dual Z-scheme g-C3N4/Fe2TiO5/Fe2O3 ternary nanocomposite using natural ilmenite for efficient photocatalysis and photosterilization under visible light
    (Elsevier, 2022-11-11) Thambiliyagodage, C; Liyanaarachchi, H; Kumar, A; Jayanetti, M; Usgodaarachchi, L; Lansakara, B
    The advanced oxidation process is a prominent method available to remove dyes released to normal water reservoirs to alleviate water scarcity. We report the fabrication of novel g-C3N4/Fe2TiO5/Fe2O3 using natural ilmenite sand as the precursor of the metallic semi-conductors exploration of a heterostructure for photodegradation of methylene blue under sunlight. Ternary composites were synthesized by varying g-C3N4 with respect to Fe2TiO5/Fe2O3 and varying Fe2TiO5/Fe2O3 with respect to g-C3N4 where the varying component was varied as 8, 24 and 40%, respect to the constant material. The hybridization of the three semi-conductors has been confirmed by the microscopic, chemical, and structural analyses. X-ray diffraction patterns show the presence of all three g-C3N4, Fe2TiO5 and α-Fe2O3 while the transmission electronic microscopic and scanning electronic microscopic images show the heterogeneous distribution of the metal oxide nanoparticles on g-C3N4 matrix forming the composite. HRTEM images further reveal the junction of Fe2TiO5 and α-Fe2O3. X-ray photoelectron spectra show the existence of s-triazine and heptazine rings in the composites with Fe3+ and Ti4+ as the only oxidation states of Fe and Ti. Fe2TiO5/Fe2O3/40% g-C3N4 with bandgap of 2.63 eV calculated by diffuse reflectance UV-Visible spectroscopy showed the highest photocatalytic activity (0.009 min−1) being 1.3 times greater than the Fe2TiO5/Fe2O3 nanoparticles. Enhanced photocatalytic activity over the fabricated composites was observed due to the increased visible light absorption, efficient charge separation and improved charge transportation. g-C3N4 coupled with 40% Fe2TiO5/Fe2O3 showed the highest antibacterial activity against gram-negative E.Coli. The synthesis of dual Z-scheme g-C3N4/Fe2TiO5/Fe2O3 ternary composite provides new sights in developing novel photocatalysts using natural ilmenite sand for environmental applications.
  • Thumbnail Image
    PublicationOpen Access
    Facial identity recognition using StyleGAN3 inversion and improved tiny YOLOv7 model
    (Nature Research, 2025-03-17) Kumar, A; Bhattacharjee, S; Kumar, Ambrish; Jayakody, D. N.K
    Facial identity recognition is one of the challenging problems in the domain of computer vision. Facial identity comprises the facial attributes of a person’s face ranging from age progression, gender, hairstyle, etc. Manipulating facial attributes such as changing the gender, hairstyle, expressions, and makeup changes the entire facial identity of a person which is often used by law offenders to commit crimes. Leveraging the deep learning-based approaches, this work proposes a one-step solution for facial attribute manipulation and detection leading to facial identity recognition in few-shot and traditional scenarios. As a first step towards performing facial identity recognition, we created the Facial Attribute Manipulation Detection (FAM) Dataset which consists of twenty unique identities with thirty-eight facial attributes generated by the StyleGAN3 inversion. The Facial Attribute Detection (FAM) Dataset has 11,560 images richly annotated in YOLO format. To perform facial attribute and identity detection, we developed the Spatial Transformer Block (STB) and Squeeze-Excite Spatial Pyramid Pooling (SE-SPP)-based Tiny YOLOv7 model and proposed as FIR-Tiny YOLOv7 (Facial Identity Recognition-Tiny YOLOv7) model. The proposed model is an improvised variant of the Tiny YOLOv7 model. For facial identity recognition, the proposed model achieved 10.0% higher mAP in the one-shot scenario, 30.4% higher mAP in the three-shot scenario, 15.3% higher mAP in the five-shot scenario, and 0.1% higher mAP in the traditional 70% − 30% split scenario as compared to the Tiny YOLOv7 model. The results obtained with the proposed model are promising for general facial identity recognition under varying facial attribute manipulation.
  • Thumbnail Image
    PublicationOpen Access
    Facial identity recognition using StyleGAN3 inversion and improved tiny YOLOv7 model
    (www.nature.com, 2025-03-17) Kumar, A; Bhattacharjee, S; Kumar, A; Jayakody, D. N. K
    Facial identity recognition is one of the challenging problems in the domain of computer vision. Facial identity comprises the facial attributes of a person’s face ranging from age progression, gender, hairstyle, etc. Manipulating facial attributes such as changing the gender, hairstyle, expressions, and makeup changes the entire facial identity of a person which is often used by law offenders to commit crimes. Leveraging the deep learning-based approaches, this work proposes a one-step solution for facial attribute manipulation and detection leading to facial identity recognition in few-shot and traditional scenarios. As a first step towards performing facial identity recognition, we created the Facial Attribute Manipulation Detection (FAM) Dataset which consists of twenty unique identities with thirty-eight facial attributes generated by the StyleGAN3 inversion. The Facial Attribute Detection (FAM) Dataset has 11,560 images richly annotated in YOLO format. To perform facial attribute and identity detection, we developed the Spatial Transformer Block (STB) and Squeeze-Excite Spatial Pyramid Pooling (SE-SPP)-based Tiny YOLOv7 model and proposed as FIR-Tiny YOLOv7 (Facial Identity Recognition-Tiny YOLOv7) model. The proposed model is an improvised variant of the Tiny YOLOv7 model. For facial identity recognition, the proposed model achieved 10.0% higher mAP in the one-shot scenario, 30.4% higher mAP in the three-shot scenario, 15.3% higher mAP in the five-shot scenario, and 0.1% higher mAP in the traditional 70% − 30% split scenario as compared to the Tiny YOLOv7 model. The results obtained with the proposed model are promising for general facial identity recognition under varying facial attribute manipulation.